3 @settitle The C Preprocessor
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1987, 1989, 1991, 1992, 1993, 1994, 1995, 1996,
12 1997, 1998, 1999, 2000, 2001, 2002
13 Free Software Foundation, Inc.
15 Permission is granted to copy, distribute and/or modify this document
16 under the terms of the GNU Free Documentation License, Version 1.1 or
17 any later version published by the Free Software Foundation. A copy of
18 the license is included in the
20 section entitled ``GNU Free Documentation License''.
22 @c man begin COPYRIGHT
29 @c man begin COPYRIGHT
30 This manual contains no Invariant Sections. The Front-Cover Texts are
31 (a) (see below), and the Back-Cover Texts are (b) (see below).
33 (a) The FSF's Front-Cover Text is:
37 (b) The FSF's Back-Cover Text is:
39 You have freedom to copy and modify this GNU Manual, like GNU
40 software. Copies published by the Free Software Foundation raise
41 funds for GNU development.
45 @macro gcctabopt{body}
49 @c Create a separate index for command line options.
53 @c Used in cppopts.texi and cppenv.texi.
57 @dircategory Programming
59 * Cpp: (cpp). The GNU C preprocessor.
64 @title The C Preprocessor
65 @subtitle Last revised April 2001
66 @subtitle for GCC version 3
67 @author Richard M. Stallman
68 @author Zachary Weinberg
70 @c There is a fill at the bottom of the page, so we need a filll to
72 @vskip 0pt plus 1filll
82 The C preprocessor implements the macro language used to transform C,
83 C++, and Objective-C programs before they are compiled. It can also be
95 * Preprocessor Output::
97 * Implementation Details::
99 * Environment Variables::
100 * GNU Free Documentation License::
101 * Index of Directives::
106 --- The Detailed Node Listing ---
110 * Initial processing::
112 * The preprocessing language::
117 * Include Operation::
119 * Once-Only Headers::
120 * Computed Includes::
126 * Object-like Macros::
127 * Function-like Macros::
132 * Predefined Macros::
133 * Undefining and Redefining Macros::
134 * Directives Within Macro Arguments::
139 * Standard Predefined Macros::
140 * Common Predefined Macros::
141 * System-specific Predefined Macros::
142 * C++ Named Operators::
147 * Operator Precedence Problems::
148 * Swallowing the Semicolon::
149 * Duplication of Side Effects::
150 * Self-Referential Macros::
152 * Newlines in Arguments::
157 * Conditional Syntax::
168 Implementation Details
170 * Implementation-defined behavior::
171 * Implementation limits::
172 * Obsolete Features::
173 * Differences from previous versions::
178 * Obsolete once-only headers::
189 @c man begin DESCRIPTION
190 The C preprocessor, often known as @dfn{cpp}, is a @dfn{macro processor}
191 that is used automatically by the C compiler to transform your program
192 before compilation. It is called a macro processor because it allows
193 you to define @dfn{macros}, which are brief abbreviations for longer
196 The C preprocessor is intended to be used only with C, C++, and
197 Objective-C source code. In the past, it has been abused as a general
198 text processor. It will choke on input which does not obey C's lexical
199 rules. For example, apostrophes will be interpreted as the beginning of
200 character constants, and cause errors. Also, you cannot rely on it
201 preserving characteristics of the input which are not significant to
202 C-family languages. If a Makefile is preprocessed, all the hard tabs
203 will be removed, and the Makefile will not work.
205 Having said that, you can often get away with using cpp on things which
206 are not C@. Other Algol-ish programming languages are often safe
207 (Pascal, Ada, etc.) So is assembly, with caution. @option{-traditional-cpp}
208 mode preserves more white space, and is otherwise more permissive. Many
209 of the problems can be avoided by writing C or C++ style comments
210 instead of native language comments, and keeping macros simple.
212 Wherever possible, you should use a preprocessor geared to the language
213 you are writing in. Modern versions of the GNU assembler have macro
214 facilities. Most high level programming languages have their own
215 conditional compilation and inclusion mechanism. If all else fails,
216 try a true general text processor, such as GNU M4.
218 C preprocessors vary in some details. This manual discusses the GNU C
219 preprocessor, which provides a small superset of the features of ISO
220 Standard C@. In its default mode, the GNU C preprocessor does not do a
221 few things required by the standard. These are features which are
222 rarely, if ever, used, and may cause surprising changes to the meaning
223 of a program which does not expect them. To get strict ISO Standard C,
224 you should use the @option{-std=c89} or @option{-std=c99} options, depending
225 on which version of the standard you want. To get all the mandatory
226 diagnostics, you must also use @option{-pedantic}. @xref{Invocation}.
228 This manual describes the behavior of the ISO preprocessor. To
229 minimize gratuitous differences, where the ISO preprocessor's
230 behavior does not conflict with traditional semantics, the
231 traditional preprocessor should behave the same way. The various
232 differences that do exist are detailed in the section @ref{Traditional
235 For clarity, unless noted otherwise, references to @samp{CPP} in this
236 manual refer to GNU CPP.
240 * Initial processing::
242 * The preprocessing language::
245 @node Initial processing
246 @section Initial processing
248 The preprocessor performs a series of textual transformations on its
249 input. These happen before all other processing. Conceptually, they
250 happen in a rigid order, and the entire file is run through each
251 transformation before the next one begins. CPP actually does them
252 all at once, for performance reasons. These transformations correspond
253 roughly to the first three ``phases of translation'' described in the C
258 @cindex character sets
260 The input file is read into memory and broken into lines.
262 CPP expects its input to be a text file, that is, an unstructured
263 stream of ASCII characters, with some characters indicating the end of a
264 line of text. Extended ASCII character sets, such as ISO Latin-1 or
265 Unicode encoded in UTF-8, are also acceptable. Character sets that are
266 not strict supersets of seven-bit ASCII will not work. We plan to add
267 complete support for international character sets in a future release.
269 Different systems use different conventions to indicate the end of a
270 line. GCC accepts the ASCII control sequences @kbd{LF}, @kbd{@w{CR
271 LF}}, @kbd{CR}, and @kbd{@w{LF CR}} as end-of-line markers. The first
272 three are the canonical sequences used by Unix, DOS and VMS, and the
273 classic Mac OS (before OSX) respectively. You may therefore safely copy
274 source code written on any of those systems to a different one and use
275 it without conversion. (GCC may lose track of the current line number
276 if a file doesn't consistently use one convention, as sometimes happens
277 when it is edited on computers with different conventions that share a
278 network file system.) @kbd{@w{LF CR}} is included because it has been
279 reported as an end-of-line marker under exotic conditions.
281 If the last line of any input file lacks an end-of-line marker, the end
282 of the file is considered to implicitly supply one. The C standard says
283 that this condition provokes undefined behavior, so GCC will emit a
288 @anchor{trigraphs}If trigraphs are enabled, they are replaced by their
289 corresponding single characters. By default GCC ignores trigraphs,
290 but if you request a strictly conforming mode with the @option{-std}
291 option, or you specify the @option{-trigraphs} option, then it
294 These are nine three-character sequences, all starting with @samp{??},
295 that are defined by ISO C to stand for single characters. They permit
296 obsolete systems that lack some of C's punctuation to use C@. For
297 example, @samp{??/} stands for @samp{\}, so @t{'??/n'} is a character
298 constant for a newline.
300 Trigraphs are not popular and many compilers implement them incorrectly.
301 Portable code should not rely on trigraphs being either converted or
302 ignored. If you use the @option{-Wall} or @option{-Wtrigraphs} options,
303 GCC will warn you when a trigraph would change the meaning of your
304 program if it were converted.
306 In a string constant, you can prevent a sequence of question marks from
307 being confused with a trigraph by inserting a backslash between the
308 question marks. @t{"(??\?)"} is the string @samp{(???)}, not
309 @samp{(?]}. Traditional C compilers do not recognize this idiom.
311 The nine trigraphs and their replacements are
314 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??-
315 Replacement: [ ] @{ @} # \ ^ | ~
319 @cindex continued lines
320 @cindex backslash-newline
321 Continued lines are merged into one long line.
323 A continued line is a line which ends with a backslash, @samp{\}. The
324 backslash is removed and the following line is joined with the current
325 one. No space is inserted, so you may split a line anywhere, even in
326 the middle of a word. (It is generally more readable to split lines
327 only at white space.)
329 The trailing backslash on a continued line is commonly referred to as a
330 @dfn{backslash-newline}.
332 If there is white space between a backslash and the end of a line, that
333 is still a continued line. However, as this is usually the result of an
334 editing mistake, and many compilers will not accept it as a continued
335 line, GCC will warn you about it.
339 @cindex line comments
340 @cindex block comments
341 All comments are replaced with single spaces.
343 There are two kinds of comments. @dfn{Block comments} begin with
344 @samp{/*} and continue until the next @samp{*/}. Block comments do not
348 /* @r{this is} /* @r{one comment} */ @r{text outside comment}
351 @dfn{Line comments} begin with @samp{//} and continue to the end of the
352 current line. Line comments do not nest either, but it does not matter,
353 because they would end in the same place anyway.
356 // @r{this is} // @r{one comment}
357 @r{text outside comment}
361 It is safe to put line comments inside block comments, or vice versa.
366 // @r{contains line comment}
368 */ @r{outside comment}
370 // @r{line comment} /* @r{contains block comment} */
374 But beware of commenting out one end of a block comment with a line
379 // @r{l.c.} /* @r{block comment begins}
380 @r{oops! this isn't a comment anymore} */
384 Comments are not recognized within string literals. @t{@w{"/* blah
385 */"}} is the string constant @samp{@w{/* blah */}}, not an empty string.
387 Line comments are not in the 1989 edition of the C standard, but they
388 are recognized by GCC as an extension. In C++ and in the 1999 edition
389 of the C standard, they are an official part of the language.
391 Since these transformations happen before all other processing, you can
392 split a line mechanically with backslash-newline anywhere. You can
393 comment out the end of a line. You can continue a line comment onto the
394 next line with backslash-newline. You can even split @samp{/*},
395 @samp{*/}, and @samp{//} onto multiple lines with backslash-newline.
411 is equivalent to @code{@w{#define FOO 1020}}. All these tricks are
412 extremely confusing and should not be used in code intended to be
415 There is no way to prevent a backslash at the end of a line from being
416 interpreted as a backslash-newline. This cannot affect any correct
420 @section Tokenization
423 @cindex preprocessing tokens
424 After the textual transformations are finished, the input file is
425 converted into a sequence of @dfn{preprocessing tokens}. These mostly
426 correspond to the syntactic tokens used by the C compiler, but there are
427 a few differences. White space separates tokens; it is not itself a
428 token of any kind. Tokens do not have to be separated by white space,
429 but it is often necessary to avoid ambiguities.
431 When faced with a sequence of characters that has more than one possible
432 tokenization, the preprocessor is greedy. It always makes each token,
433 starting from the left, as big as possible before moving on to the next
434 token. For instance, @code{a+++++b} is interpreted as
435 @code{@w{a ++ ++ + b}}, not as @code{@w{a ++ + ++ b}}, even though the
436 latter tokenization could be part of a valid C program and the former
439 Once the input file is broken into tokens, the token boundaries never
440 change, except when the @samp{##} preprocessing operator is used to paste
441 tokens together. @xref{Concatenation}. For example,
453 The compiler does not re-tokenize the preprocessor's output. Each
454 preprocessing token becomes one compiler token.
457 Preprocessing tokens fall into five broad classes: identifiers,
458 preprocessing numbers, string literals, punctuators, and other. An
459 @dfn{identifier} is the same as an identifier in C: any sequence of
460 letters, digits, or underscores, which begins with a letter or
461 underscore. Keywords of C have no significance to the preprocessor;
462 they are ordinary identifiers. You can define a macro whose name is a
463 keyword, for instance. The only identifier which can be considered a
464 preprocessing keyword is @code{defined}. @xref{Defined}.
466 This is mostly true of other languages which use the C preprocessor.
467 However, a few of the keywords of C++ are significant even in the
468 preprocessor. @xref{C++ Named Operators}.
470 In the 1999 C standard, identifiers may contain letters which are not
471 part of the ``basic source character set,'' at the implementation's
472 discretion (such as accented Latin letters, Greek letters, or Chinese
473 ideograms). This may be done with an extended character set, or the
474 @samp{\u} and @samp{\U} escape sequences. GCC does not presently
475 implement either feature in the preprocessor or the compiler.
477 As an extension, GCC treats @samp{$} as a letter. This is for
478 compatibility with some systems, such as VMS, where @samp{$} is commonly
479 used in system-defined function and object names. @samp{$} is not a
480 letter in strictly conforming mode, or if you specify the @option{-$}
481 option. @xref{Invocation}.
484 @cindex preprocessing numbers
485 A @dfn{preprocessing number} has a rather bizarre definition. The
486 category includes all the normal integer and floating point constants
487 one expects of C, but also a number of other things one might not
488 initially recognize as a number. Formally, preprocessing numbers begin
489 with an optional period, a required decimal digit, and then continue
490 with any sequence of letters, digits, underscores, periods, and
491 exponents. Exponents are the two-character sequences @samp{e+},
492 @samp{e-}, @samp{E+}, @samp{E-}, @samp{p+}, @samp{p-}, @samp{P+}, and
493 @samp{P-}. (The exponents that begin with @samp{p} or @samp{P} are new
494 to C99. They are used for hexadecimal floating-point constants.)
496 The purpose of this unusual definition is to isolate the preprocessor
497 from the full complexity of numeric constants. It does not have to
498 distinguish between lexically valid and invalid floating-point numbers,
499 which is complicated. The definition also permits you to split an
500 identifier at any position and get exactly two tokens, which can then be
501 pasted back together with the @samp{##} operator.
503 It's possible for preprocessing numbers to cause programs to be
504 misinterpreted. For example, @code{0xE+12} is a preprocessing number
505 which does not translate to any valid numeric constant, therefore a
506 syntax error. It does not mean @code{@w{0xE + 12}}, which is what you
509 @cindex string literals
510 @cindex string constants
511 @cindex character constants
512 @cindex header file names
513 @c the @: prevents makeinfo from turning '' into ".
514 @dfn{String literals} are string constants, character constants, and
515 header file names (the argument of @samp{#include}).@footnote{The C
516 standard uses the term @dfn{string literal} to refer only to what we are
517 calling @dfn{string constants}.} String constants and character
518 constants are straightforward: @t{"@dots{}"} or @t{'@dots{}'}. In
519 either case embedded quotes should be escaped with a backslash:
520 @t{'\'@:'} is the character constant for @samp{'}. There is no limit on
521 the length of a character constant, but the value of a character
522 constant that contains more than one character is
523 implementation-defined. @xref{Implementation Details}.
525 Header file names either look like string constants, @t{"@dots{}"}, or are
526 written with angle brackets instead, @t{<@dots{}>}. In either case,
527 backslash is an ordinary character. There is no way to escape the
528 closing quote or angle bracket. The preprocessor looks for the header
529 file in different places depending on which form you use. @xref{Include
532 No string literal may extend past the end of a line. Older versions
533 of GCC accepted multi-line string constants. You may use continued
534 lines instead, or string constant concatenation. @xref{Differences
535 from previous versions}.
539 @cindex alternative tokens
540 @dfn{Punctuators} are all the usual bits of punctuation which are
541 meaningful to C and C++. All but three of the punctuation characters in
542 ASCII are C punctuators. The exceptions are @samp{@@}, @samp{$}, and
543 @samp{`}. In addition, all the two- and three-character operators are
544 punctuators. There are also six @dfn{digraphs}, which the C++ standard
545 calls @dfn{alternative tokens}, which are merely alternate ways to spell
546 other punctuators. This is a second attempt to work around missing
547 punctuation in obsolete systems. It has no negative side effects,
548 unlike trigraphs, but does not cover as much ground. The digraphs and
549 their corresponding normal punctuators are:
552 Digraph: <% %> <: :> %: %:%:
553 Punctuator: @{ @} [ ] # ##
557 Any other single character is considered ``other.'' It is passed on to
558 the preprocessor's output unmolested. The C compiler will almost
559 certainly reject source code containing ``other'' tokens. In ASCII, the
560 only other characters are @samp{@@}, @samp{$}, @samp{`}, and control
561 characters other than NUL (all bits zero). (Note that @samp{$} is
562 normally considered a letter.) All characters with the high bit set
563 (numeric range 0x7F--0xFF) are also ``other'' in the present
564 implementation. This will change when proper support for international
565 character sets is added to GCC@.
567 NUL is a special case because of the high probability that its
568 appearance is accidental, and because it may be invisible to the user
569 (many terminals do not display NUL at all). Within comments, NULs are
570 silently ignored, just as any other character would be. In running
571 text, NUL is considered white space. For example, these two directives
572 have the same meaning.
580 (where @samp{^@@} is ASCII NUL)@. Within string or character constants,
581 NULs are preserved. In the latter two cases the preprocessor emits a
584 @node The preprocessing language
585 @section The preprocessing language
587 @cindex preprocessing directives
588 @cindex directive line
589 @cindex directive name
591 After tokenization, the stream of tokens may simply be passed straight
592 to the compiler's parser. However, if it contains any operations in the
593 @dfn{preprocessing language}, it will be transformed first. This stage
594 corresponds roughly to the standard's ``translation phase 4'' and is
595 what most people think of as the preprocessor's job.
597 The preprocessing language consists of @dfn{directives} to be executed
598 and @dfn{macros} to be expanded. Its primary capabilities are:
602 Inclusion of header files. These are files of declarations that can be
603 substituted into your program.
606 Macro expansion. You can define @dfn{macros}, which are abbreviations
607 for arbitrary fragments of C code. The preprocessor will replace the
608 macros with their definitions throughout the program. Some macros are
609 automatically defined for you.
612 Conditional compilation. You can include or exclude parts of the
613 program according to various conditions.
616 Line control. If you use a program to combine or rearrange source files
617 into an intermediate file which is then compiled, you can use line
618 control to inform the compiler where each source line originally came
622 Diagnostics. You can detect problems at compile time and issue errors
626 There are a few more, less useful, features.
628 Except for expansion of predefined macros, all these operations are
629 triggered with @dfn{preprocessing directives}. Preprocessing directives
630 are lines in your program that start with @samp{#}. Whitespace is
631 allowed before and after the @samp{#}. The @samp{#} is followed by an
632 identifier, the @dfn{directive name}. It specifies the operation to
633 perform. Directives are commonly referred to as @samp{#@var{name}}
634 where @var{name} is the directive name. For example, @samp{#define} is
635 the directive that defines a macro.
637 The @samp{#} which begins a directive cannot come from a macro
638 expansion. Also, the directive name is not macro expanded. Thus, if
639 @code{foo} is defined as a macro expanding to @code{define}, that does
640 not make @samp{#foo} a valid preprocessing directive.
642 The set of valid directive names is fixed. Programs cannot define new
643 preprocessing directives.
645 Some directives require arguments; these make up the rest of the
646 directive line and must be separated from the directive name by
647 whitespace. For example, @samp{#define} must be followed by a macro
648 name and the intended expansion of the macro.
650 A preprocessing directive cannot cover more than one line. The line
651 may, however, be continued with backslash-newline, or by a block comment
652 which extends past the end of the line. In either case, when the
653 directive is processed, the continuations have already been merged with
654 the first line to make one long line.
657 @chapter Header Files
660 A header file is a file containing C declarations and macro definitions
661 (@pxref{Macros}) to be shared between several source files. You request
662 the use of a header file in your program by @dfn{including} it, with the
663 C preprocessing directive @samp{#include}.
665 Header files serve two purposes.
669 @cindex system header files
670 System header files declare the interfaces to parts of the operating
671 system. You include them in your program to supply the definitions and
672 declarations you need to invoke system calls and libraries.
675 Your own header files contain declarations for interfaces between the
676 source files of your program. Each time you have a group of related
677 declarations and macro definitions all or most of which are needed in
678 several different source files, it is a good idea to create a header
682 Including a header file produces the same results as copying the header
683 file into each source file that needs it. Such copying would be
684 time-consuming and error-prone. With a header file, the related
685 declarations appear in only one place. If they need to be changed, they
686 can be changed in one place, and programs that include the header file
687 will automatically use the new version when next recompiled. The header
688 file eliminates the labor of finding and changing all the copies as well
689 as the risk that a failure to find one copy will result in
690 inconsistencies within a program.
692 In C, the usual convention is to give header files names that end with
693 @file{.h}. It is most portable to use only letters, digits, dashes, and
694 underscores in header file names, and at most one dot.
698 * Include Operation::
700 * Once-Only Headers::
701 * Computed Includes::
707 @section Include Syntax
710 Both user and system header files are included using the preprocessing
711 directive @samp{#include}. It has two variants:
714 @item #include <@var{file}>
715 This variant is used for system header files. It searches for a file
716 named @var{file} in a standard list of system directories. You can prepend
717 directories to this list with the @option{-I} option (@pxref{Invocation}).
719 @item #include "@var{file}"
720 This variant is used for header files of your own program. It searches
721 for a file named @var{file} first in the directory containing the
722 current file, then in the same directories used for @code{<@var{file}>}.
725 The argument of @samp{#include}, whether delimited with quote marks or
726 angle brackets, behaves like a string constant in that comments are not
727 recognized, and macro names are not expanded. Thus, @code{@w{#include
728 <x/*y>}} specifies inclusion of a system header file named @file{x/*y}.
730 However, if backslashes occur within @var{file}, they are considered
731 ordinary text characters, not escape characters. None of the character
732 escape sequences appropriate to string constants in C are processed.
733 Thus, @code{@w{#include "x\n\\y"}} specifies a filename containing three
734 backslashes. (Some systems interpret @samp{\} as a pathname separator.
735 All of these also interpret @samp{/} the same way. It is most portable
736 to use only @samp{/}.)
738 It is an error if there is anything (other than comments) on the line
741 @node Include Operation
742 @section Include Operation
744 The @samp{#include} directive works by directing the C preprocessor to
745 scan the specified file as input before continuing with the rest of the
746 current file. The output from the preprocessor contains the output
747 already generated, followed by the output resulting from the included
748 file, followed by the output that comes from the text after the
749 @samp{#include} directive. For example, if you have a header file
750 @file{header.h} as follows,
757 and a main program called @file{program.c} that uses the header file,
772 the compiler will see the same token stream as it would if
773 @file{program.c} read
786 Included files are not limited to declarations and macro definitions;
787 those are merely the typical uses. Any fragment of a C program can be
788 included from another file. The include file could even contain the
789 beginning of a statement that is concluded in the containing file, or
790 the end of a statement that was started in the including file. However,
791 an included file must consist of complete tokens. Comments and string
792 literals which have not been closed by the end of an included file are
793 invalid. For error recovery, they are considered to end at the end of
796 To avoid confusion, it is best if header files contain only complete
797 syntactic units---function declarations or definitions, type
800 The line following the @samp{#include} directive is always treated as a
801 separate line by the C preprocessor, even if the included file lacks a
807 GCC looks in several different places for headers. On a normal Unix
808 system, if you do not instruct it otherwise, it will look for headers
809 requested with @code{@w{#include <@var{file}>}} in:
813 /usr/lib/gcc-lib/@var{target}/@var{version}/include
814 /usr/@var{target}/include
818 For C++ programs, it will also look in @file{/usr/include/g++-v3},
819 first. In the above, @var{target} is the canonical name of the system
820 GCC was configured to compile code for; often but not always the same as
821 the canonical name of the system it runs on. @var{version} is the
822 version of GCC in use.
824 You can add to this list with the @option{-I@var{dir}} command line
825 option. All the directories named by @option{-I} are searched, in
826 left-to-right order, @emph{before} the default directories. You can
827 also prevent GCC from searching any of the default directories with the
828 @option{-nostdinc} option. This is useful when you are compiling an
829 operating system kernel or some other program that does not use the
830 standard C library facilities, or the standard C library itself.
832 GCC looks for headers requested with @code{@w{#include "@var{file}"}}
833 first in the directory containing the current file, then in the same
834 places it would have looked for a header requested with angle brackets.
835 For example, if @file{/usr/include/sys/stat.h} contains
836 @code{@w{#include "types.h"}}, GCC looks for @file{types.h} first in
837 @file{/usr/include/sys}, then in its usual search path.
839 If you name a search directory with @option{-I@var{dir}} that is also a
840 system include directory, the @option{-I} wins; the directory will be
841 searched according to the @option{-I} ordering, and it will not be
842 treated as a system include directory. GCC will warn you when a system
843 include directory is hidden in this way.
845 @samp{#line} (@pxref{Line Control}) does not change GCC's idea of the
846 directory containing the current file.
848 You may put @option{-I-} at any point in your list of @option{-I} options.
849 This has two effects. First, directories appearing before the
850 @option{-I-} in the list are searched only for headers requested with
851 quote marks. Directories after @option{-I-} are searched for all
852 headers. Second, the directory containing the current file is not
853 searched for anything, unless it happens to be one of the directories
854 named by an @option{-I} switch.
856 @option{-I. -I-} is not the same as no @option{-I} options at all, and does
857 not cause the same behavior for @samp{<>} includes that @samp{""}
858 includes get with no special options. @option{-I.} searches the
859 compiler's current working directory for header files. That may or may
860 not be the same as the directory containing the current file.
862 If you need to look for headers in a directory named @file{-}, write
865 There are several more ways to adjust the header search path. They are
866 generally less useful. @xref{Invocation}.
868 @node Once-Only Headers
869 @section Once-Only Headers
870 @cindex repeated inclusion
871 @cindex including just once
872 @cindex wrapper @code{#ifndef}
874 If a header file happens to be included twice, the compiler will process
875 its contents twice. This is very likely to cause an error, e.g.@: when the
876 compiler sees the same structure definition twice. Even if it does not,
877 it will certainly waste time.
879 The standard way to prevent this is to enclose the entire real contents
880 of the file in a conditional, like this:
885 #ifndef FILE_FOO_SEEN
886 #define FILE_FOO_SEEN
888 @var{the entire file}
890 #endif /* !FILE_FOO_SEEN */
894 This construct is commonly known as a @dfn{wrapper #ifndef}.
895 When the header is included again, the conditional will be false,
896 because @code{FILE_FOO_SEEN} is defined. The preprocessor will skip
897 over the entire contents of the file, and the compiler will not see it
900 CPP optimizes even further. It remembers when a header file has a
901 wrapper @samp{#ifndef}. If a subsequent @samp{#include} specifies that
902 header, and the macro in the @samp{#ifndef} is still defined, it does
903 not bother to rescan the file at all.
905 You can put comments outside the wrapper. They will not interfere with
908 @cindex controlling macro
910 The macro @code{FILE_FOO_SEEN} is called the @dfn{controlling macro} or
911 @dfn{guard macro}. In a user header file, the macro name should not
912 begin with @samp{_}. In a system header file, it should begin with
913 @samp{__} to avoid conflicts with user programs. In any kind of header
914 file, the macro name should contain the name of the file and some
915 additional text, to avoid conflicts with other header files.
917 @node Computed Includes
918 @section Computed Includes
919 @cindex computed includes
920 @cindex macros in include
922 Sometimes it is necessary to select one of several different header
923 files to be included into your program. They might specify
924 configuration parameters to be used on different sorts of operating
925 systems, for instance. You could do this with a series of conditionals,
929 # include "system_1.h"
931 # include "system_2.h"
937 That rapidly becomes tedious. Instead, the preprocessor offers the
938 ability to use a macro for the header name. This is called a
939 @dfn{computed include}. Instead of writing a header name as the direct
940 argument of @samp{#include}, you simply put a macro name there instead:
943 #define SYSTEM_H "system_1.h"
949 @code{SYSTEM_H} will be expanded, and the preprocessor will look for
950 @file{system_1.h} as if the @samp{#include} had been written that way
951 originally. @code{SYSTEM_H} could be defined by your Makefile with a
954 You must be careful when you define the macro. @samp{#define} saves
955 tokens, not text. The preprocessor has no way of knowing that the macro
956 will be used as the argument of @samp{#include}, so it generates
957 ordinary tokens, not a header name. This is unlikely to cause problems
958 if you use double-quote includes, which are close enough to string
959 constants. If you use angle brackets, however, you may have trouble.
961 The syntax of a computed include is actually a bit more general than the
962 above. If the first non-whitespace character after @samp{#include} is
963 not @samp{"} or @samp{<}, then the entire line is macro-expanded
964 like running text would be.
966 If the line expands to a single string constant, the contents of that
967 string constant are the file to be included. CPP does not re-examine the
968 string for embedded quotes, but neither does it process backslash
969 escapes in the string. Therefore
972 #define HEADER "a\"b"
977 looks for a file named @file{a\"b}. CPP searches for the file according
978 to the rules for double-quoted includes.
980 If the line expands to a token stream beginning with a @samp{<} token
981 and including a @samp{>} token, then the tokens between the @samp{<} and
982 the first @samp{>} are combined to form the filename to be included.
983 Any whitespace between tokens is reduced to a single space; then any
984 space after the initial @samp{<} is retained, but a trailing space
985 before the closing @samp{>} is ignored. CPP searches for the file
986 according to the rules for angle-bracket includes.
988 In either case, if there are any tokens on the line after the file name,
989 an error occurs and the directive is not processed. It is also an error
990 if the result of expansion does not match either of the two expected
993 These rules are implementation-defined behavior according to the C
994 standard. To minimize the risk of different compilers interpreting your
995 computed includes differently, we recommend you use only a single
996 object-like macro which expands to a string constant. This will also
997 minimize confusion for people reading your program.
999 @node Wrapper Headers
1000 @section Wrapper Headers
1001 @cindex wrapper headers
1002 @cindex overriding a header file
1003 @findex #include_next
1005 Sometimes it is necessary to adjust the contents of a system-provided
1006 header file without editing it directly. GCC's @command{fixincludes}
1007 operation does this, for example. One way to do that would be to create
1008 a new header file with the same name and insert it in the search path
1009 before the original header. That works fine as long as you're willing
1010 to replace the old header entirely. But what if you want to refer to
1011 the old header from the new one?
1013 You cannot simply include the old header with @samp{#include}. That
1014 will start from the beginning, and find your new header again. If your
1015 header is not protected from multiple inclusion (@pxref{Once-Only
1016 Headers}), it will recurse infinitely and cause a fatal error.
1018 You could include the old header with an absolute pathname:
1020 #include "/usr/include/old-header.h"
1023 This works, but is not clean; should the system headers ever move, you
1024 would have to edit the new headers to match.
1026 There is no way to solve this problem within the C standard, but you can
1027 use the GNU extension @samp{#include_next}. It means, ``Include the
1028 @emph{next} file with this name.'' This directive works like
1029 @samp{#include} except in searching for the specified file: it starts
1030 searching the list of header file directories @emph{after} the directory
1031 in which the current file was found.
1033 Suppose you specify @option{-I /usr/local/include}, and the list of
1034 directories to search also includes @file{/usr/include}; and suppose
1035 both directories contain @file{signal.h}. Ordinary @code{@w{#include
1036 <signal.h>}} finds the file under @file{/usr/local/include}. If that
1037 file contains @code{@w{#include_next <signal.h>}}, it starts searching
1038 after that directory, and finds the file in @file{/usr/include}.
1040 @samp{#include_next} does not distinguish between @code{<@var{file}>}
1041 and @code{"@var{file}"} inclusion, nor does it check that the file you
1042 specify has the same name as the current file. It simply looks for the
1043 file named, starting with the directory in the search path after the one
1044 where the current file was found.
1046 The use of @samp{#include_next} can lead to great confusion. We
1047 recommend it be used only when there is no other alternative. In
1048 particular, it should not be used in the headers belonging to a specific
1049 program; it should be used only to make global corrections along the
1050 lines of @command{fixincludes}.
1052 @node System Headers
1053 @section System Headers
1054 @cindex system header files
1056 The header files declaring interfaces to the operating system and
1057 runtime libraries often cannot be written in strictly conforming C@.
1058 Therefore, GCC gives code found in @dfn{system headers} special
1059 treatment. All warnings, other than those generated by @samp{#warning}
1060 (@pxref{Diagnostics}), are suppressed while GCC is processing a system
1061 header. Macros defined in a system header are immune to a few warnings
1062 wherever they are expanded. This immunity is granted on an ad-hoc
1063 basis, when we find that a warning generates lots of false positives
1064 because of code in macros defined in system headers.
1066 Normally, only the headers found in specific directories are considered
1067 system headers. These directories are determined when GCC is compiled.
1068 There are, however, two ways to make normal headers into system headers.
1070 The @option{-isystem} command line option adds its argument to the list of
1071 directories to search for headers, just like @option{-I}. Any headers
1072 found in that directory will be considered system headers.
1074 All directories named by @option{-isystem} are searched @emph{after} all
1075 directories named by @option{-I}, no matter what their order was on the
1076 command line. If the same directory is named by both @option{-I} and
1077 @option{-isystem}, @option{-I} wins; it is as if the @option{-isystem} option
1078 had never been specified at all. GCC warns you when this happens.
1080 @findex #pragma GCC system_header
1081 There is also a directive, @code{@w{#pragma GCC system_header}}, which
1082 tells GCC to consider the rest of the current include file a system
1083 header, no matter where it was found. Code that comes before the
1084 @samp{#pragma} in the file will not be affected. @code{@w{#pragma GCC
1085 system_header}} has no effect in the primary source file.
1087 On very old systems, some of the pre-defined system header directories
1088 get even more special treatment. GNU C++ considers code in headers
1089 found in those directories to be surrounded by an @code{@w{extern "C"}}
1090 block. There is no way to request this behavior with a @samp{#pragma},
1091 or from the command line.
1096 A @dfn{macro} is a fragment of code which has been given a name.
1097 Whenever the name is used, it is replaced by the contents of the macro.
1098 There are two kinds of macros. They differ mostly in what they look
1099 like when they are used. @dfn{Object-like} macros resemble data objects
1100 when used, @dfn{function-like} macros resemble function calls.
1102 You may define any valid identifier as a macro, even if it is a C
1103 keyword. The preprocessor does not know anything about keywords. This
1104 can be useful if you wish to hide a keyword such as @code{const} from an
1105 older compiler that does not understand it. However, the preprocessor
1106 operator @code{defined} (@pxref{Defined}) can never be defined as a
1107 macro, and C++'s named operators (@pxref{C++ Named Operators}) cannot be
1108 macros when you are compiling C++.
1111 * Object-like Macros::
1112 * Function-like Macros::
1117 * Predefined Macros::
1118 * Undefining and Redefining Macros::
1119 * Directives Within Macro Arguments::
1123 @node Object-like Macros
1124 @section Object-like Macros
1125 @cindex object-like macro
1126 @cindex symbolic constants
1127 @cindex manifest constants
1129 An @dfn{object-like macro} is a simple identifier which will be replaced
1130 by a code fragment. It is called object-like because it looks like a
1131 data object in code that uses it. They are most commonly used to give
1132 symbolic names to numeric constants.
1135 You create macros with the @samp{#define} directive. @samp{#define} is
1136 followed by the name of the macro and then the token sequence it should
1137 be an abbreviation for, which is variously referred to as the macro's
1138 @dfn{body}, @dfn{expansion} or @dfn{replacement list}. For example,
1141 #define BUFFER_SIZE 1024
1145 defines a macro named @code{BUFFER_SIZE} as an abbreviation for the
1146 token @code{1024}. If somewhere after this @samp{#define} directive
1147 there comes a C statement of the form
1150 foo = (char *) malloc (BUFFER_SIZE);
1154 then the C preprocessor will recognize and @dfn{expand} the macro
1155 @code{BUFFER_SIZE}. The C compiler will see the same tokens as it would
1159 foo = (char *) malloc (1024);
1162 By convention, macro names are written in upper case. Programs are
1163 easier to read when it is possible to tell at a glance which names are
1166 The macro's body ends at the end of the @samp{#define} line. You may
1167 continue the definition onto multiple lines, if necessary, using
1168 backslash-newline. When the macro is expanded, however, it will all
1169 come out on one line. For example,
1172 #define NUMBERS 1, \
1175 int x[] = @{ NUMBERS @};
1176 @expansion{} int x[] = @{ 1, 2, 3 @};
1180 The most common visible consequence of this is surprising line numbers
1183 There is no restriction on what can go in a macro body provided it
1184 decomposes into valid preprocessing tokens. Parentheses need not
1185 balance, and the body need not resemble valid C code. (If it does not,
1186 you may get error messages from the C compiler when you use the macro.)
1188 The C preprocessor scans your program sequentially. Macro definitions
1189 take effect at the place you write them. Therefore, the following input
1190 to the C preprocessor
1206 When the preprocessor expands a macro name, the macro's expansion
1207 replaces the macro invocation, then the expansion is examined for more
1208 macros to expand. For example,
1212 #define TABLESIZE BUFSIZE
1213 #define BUFSIZE 1024
1215 @expansion{} BUFSIZE
1221 @code{TABLESIZE} is expanded first to produce @code{BUFSIZE}, then that
1222 macro is expanded to produce the final result, @code{1024}.
1224 Notice that @code{BUFSIZE} was not defined when @code{TABLESIZE} was
1225 defined. The @samp{#define} for @code{TABLESIZE} uses exactly the
1226 expansion you specify---in this case, @code{BUFSIZE}---and does not
1227 check to see whether it too contains macro names. Only when you
1228 @emph{use} @code{TABLESIZE} is the result of its expansion scanned for
1231 This makes a difference if you change the definition of @code{BUFSIZE}
1232 at some point in the source file. @code{TABLESIZE}, defined as shown,
1233 will always expand using the definition of @code{BUFSIZE} that is
1234 currently in effect:
1237 #define BUFSIZE 1020
1238 #define TABLESIZE BUFSIZE
1244 Now @code{TABLESIZE} expands (in two stages) to @code{37}.
1246 If the expansion of a macro contains its own name, either directly or
1247 via intermediate macros, it is not expanded again when the expansion is
1248 examined for more macros. This prevents infinite recursion.
1249 @xref{Self-Referential Macros}, for the precise details.
1251 @node Function-like Macros
1252 @section Function-like Macros
1253 @cindex function-like macros
1255 You can also define macros whose use looks like a function call. These
1256 are called @dfn{function-like macros}. To define a function-like macro,
1257 you use the same @samp{#define} directive, but you put a pair of
1258 parentheses immediately after the macro name. For example,
1261 #define lang_init() c_init()
1263 @expansion{} c_init()
1266 A function-like macro is only expanded if its name appears with a pair
1267 of parentheses after it. If you write just the name, it is left alone.
1268 This can be useful when you have a function and a macro of the same
1269 name, and you wish to use the function sometimes.
1272 extern void foo(void);
1273 #define foo() /* optimized inline version */
1279 Here the call to @code{foo()} will use the macro, but the function
1280 pointer will get the address of the real function. If the macro were to
1281 be expanded, it would cause a syntax error.
1283 If you put spaces between the macro name and the parentheses in the
1284 macro definition, that does not define a function-like macro, it defines
1285 an object-like macro whose expansion happens to begin with a pair of
1289 #define lang_init () c_init()
1291 @expansion{} () c_init()()
1294 The first two pairs of parentheses in this expansion come from the
1295 macro. The third is the pair that was originally after the macro
1296 invocation. Since @code{lang_init} is an object-like macro, it does not
1297 consume those parentheses.
1299 @node Macro Arguments
1300 @section Macro Arguments
1302 @cindex macros with arguments
1303 @cindex arguments in macro definitions
1305 Function-like macros can take @dfn{arguments}, just like true functions.
1306 To define a macro that uses arguments, you insert @dfn{parameters}
1307 between the pair of parentheses in the macro definition that make the
1308 macro function-like. The parameters must be valid C identifiers,
1309 separated by commas and optionally whitespace.
1311 To invoke a macro that takes arguments, you write the name of the macro
1312 followed by a list of @dfn{actual arguments} in parentheses, separated
1313 by commas. The invocation of the macro need not be restricted to a
1314 single logical line---it can cross as many lines in the source file as
1315 you wish. The number of arguments you give must match the number of
1316 parameters in the macro definition. When the macro is expanded, each
1317 use of a parameter in its body is replaced by the tokens of the
1318 corresponding argument. (You need not use all of the parameters in the
1321 As an example, here is a macro that computes the minimum of two numeric
1322 values, as it is defined in many C programs, and some uses.
1325 #define min(X, Y) ((X) < (Y) ? (X) : (Y))
1326 x = min(a, b); @expansion{} x = ((a) < (b) ? (a) : (b));
1327 y = min(1, 2); @expansion{} y = ((1) < (2) ? (1) : (2));
1328 z = min(a + 28, *p); @expansion{} z = ((a + 28) < (*p) ? (a + 28) : (*p));
1332 (In this small example you can already see several of the dangers of
1333 macro arguments. @xref{Macro Pitfalls}, for detailed explanations.)
1335 Leading and trailing whitespace in each argument is dropped, and all
1336 whitespace between the tokens of an argument is reduced to a single
1337 space. Parentheses within each argument must balance; a comma within
1338 such parentheses does not end the argument. However, there is no
1339 requirement for square brackets or braces to balance, and they do not
1340 prevent a comma from separating arguments. Thus,
1343 macro (array[x = y, x + 1])
1347 passes two arguments to @code{macro}: @code{array[x = y} and @code{x +
1348 1]}. If you want to supply @code{array[x = y, x + 1]} as an argument,
1349 you can write it as @code{array[(x = y, x + 1)]}, which is equivalent C
1352 All arguments to a macro are completely macro-expanded before they are
1353 substituted into the macro body. After substitution, the complete text
1354 is scanned again for macros to expand, including the arguments. This rule
1355 may seem strange, but it is carefully designed so you need not worry
1356 about whether any function call is actually a macro invocation. You can
1357 run into trouble if you try to be too clever, though. @xref{Argument
1358 Prescan}, for detailed discussion.
1360 For example, @code{min (min (a, b), c)} is first expanded to
1363 min (((a) < (b) ? (a) : (b)), (c))
1371 ((((a) < (b) ? (a) : (b))) < (c)
1372 ? (((a) < (b) ? (a) : (b)))
1378 (Line breaks shown here for clarity would not actually be generated.)
1380 @cindex empty macro arguments
1381 You can leave macro arguments empty; this is not an error to the
1382 preprocessor (but many macros will then expand to invalid code).
1383 You cannot leave out arguments entirely; if a macro takes two arguments,
1384 there must be exactly one comma at the top level of its argument list.
1385 Here are some silly examples using @code{min}:
1388 min(, b) @expansion{} (( ) < (b) ? ( ) : (b))
1389 min(a, ) @expansion{} ((a ) < ( ) ? (a ) : ( ))
1390 min(,) @expansion{} (( ) < ( ) ? ( ) : ( ))
1391 min((,),) @expansion{} (((,)) < ( ) ? ((,)) : ( ))
1393 min() @error{} macro "min" requires 2 arguments, but only 1 given
1394 min(,,) @error{} macro "min" passed 3 arguments, but takes just 2
1397 Whitespace is not a preprocessing token, so if a macro @code{foo} takes
1398 one argument, @code{@w{foo ()}} and @code{@w{foo ( )}} both supply it an
1399 empty argument. Previous GNU preprocessor implementations and
1400 documentation were incorrect on this point, insisting that a
1401 function-like macro that takes a single argument be passed a space if an
1402 empty argument was required.
1404 Macro parameters appearing inside string literals are not replaced by
1405 their corresponding actual arguments.
1408 #define foo(x) x, "x"
1409 foo(bar) @expansion{} bar, "x"
1412 @node Stringification
1413 @section Stringification
1414 @cindex stringification
1415 @cindex @samp{#} operator
1417 Sometimes you may want to convert a macro argument into a string
1418 constant. Parameters are not replaced inside string constants, but you
1419 can use the @samp{#} preprocessing operator instead. When a macro
1420 parameter is used with a leading @samp{#}, the preprocessor replaces it
1421 with the literal text of the actual argument, converted to a string
1422 constant. Unlike normal parameter replacement, the argument is not
1423 macro-expanded first. This is called @dfn{stringification}.
1425 There is no way to combine an argument with surrounding text and
1426 stringify it all together. Instead, you can write a series of adjacent
1427 string constants and stringified arguments. The preprocessor will
1428 replace the stringified arguments with string constants. The C
1429 compiler will then combine all the adjacent string constants into one
1432 Here is an example of a macro definition that uses stringification:
1436 #define WARN_IF(EXP) \
1438 fprintf (stderr, "Warning: " #EXP "\n"); @} \
1441 @expansion{} do @{ if (x == 0)
1442 fprintf (stderr, "Warning: " "x == 0" "\n"); @} while (0);
1447 The argument for @code{EXP} is substituted once, as-is, into the
1448 @code{if} statement, and once, stringified, into the argument to
1449 @code{fprintf}. If @code{x} were a macro, it would be expanded in the
1450 @code{if} statement, but not in the string.
1452 The @code{do} and @code{while (0)} are a kludge to make it possible to
1453 write @code{WARN_IF (@var{arg});}, which the resemblance of
1454 @code{WARN_IF} to a function would make C programmers want to do; see
1455 @ref{Swallowing the Semicolon}.
1457 Stringification in C involves more than putting double-quote characters
1458 around the fragment. The preprocessor backslash-escapes the quotes
1459 surrounding embedded string constants, and all backslashes within string and
1460 character constants, in order to get a valid C string constant with the
1461 proper contents. Thus, stringifying @code{@w{p = "foo\n";}} results in
1462 @t{@w{"p = \"foo\\n\";"}}. However, backslashes that are not inside string
1463 or character constants are not duplicated: @samp{\n} by itself
1464 stringifies to @t{"\n"}.
1466 All leading and trailing whitespace in text being stringified is
1467 ignored. Any sequence of whitespace in the middle of the text is
1468 converted to a single space in the stringified result. Comments are
1469 replaced by whitespace long before stringification happens, so they
1470 never appear in stringified text.
1472 There is no way to convert a macro argument into a character constant.
1474 If you want to stringify the result of expansion of a macro argument,
1475 you have to use two levels of macros.
1478 #define xstr(s) str(s)
1484 @expansion{} xstr (4)
1485 @expansion{} str (4)
1489 @code{s} is stringified when it is used in @code{str}, so it is not
1490 macro-expanded first. But @code{s} is an ordinary argument to
1491 @code{xstr}, so it is completely macro-expanded before @code{xstr}
1492 itself is expanded (@pxref{Argument Prescan}). Therefore, by the time
1493 @code{str} gets to its argument, it has already been macro-expanded.
1496 @section Concatenation
1497 @cindex concatenation
1498 @cindex token pasting
1499 @cindex token concatenation
1500 @cindex @samp{##} operator
1502 It is often useful to merge two tokens into one while expanding macros.
1503 This is called @dfn{token pasting} or @dfn{token concatenation}. The
1504 @samp{##} preprocessing operator performs token pasting. When a macro
1505 is expanded, the two tokens on either side of each @samp{##} operator
1506 are combined into a single token, which then replaces the @samp{##} and
1507 the two original tokens in the macro expansion. Usually both will be
1508 identifiers, or one will be an identifier and the other a preprocessing
1509 number. When pasted, they make a longer identifier. This isn't the
1510 only valid case. It is also possible to concatenate two numbers (or a
1511 number and a name, such as @code{1.5} and @code{e3}) into a number.
1512 Also, multi-character operators such as @code{+=} can be formed by
1515 However, two tokens that don't together form a valid token cannot be
1516 pasted together. For example, you cannot concatenate @code{x} with
1517 @code{+} in either order. If you try, the preprocessor issues a warning
1518 and emits the two tokens. Whether it puts white space between the
1519 tokens is undefined. It is common to find unnecessary uses of @samp{##}
1520 in complex macros. If you get this warning, it is likely that you can
1521 simply remove the @samp{##}.
1523 Both the tokens combined by @samp{##} could come from the macro body,
1524 but you could just as well write them as one token in the first place.
1525 Token pasting is most useful when one or both of the tokens comes from a
1526 macro argument. If either of the tokens next to an @samp{##} is a
1527 parameter name, it is replaced by its actual argument before @samp{##}
1528 executes. As with stringification, the actual argument is not
1529 macro-expanded first. If the argument is empty, that @samp{##} has no
1532 Keep in mind that the C preprocessor converts comments to whitespace
1533 before macros are even considered. Therefore, you cannot create a
1534 comment by concatenating @samp{/} and @samp{*}. You can put as much
1535 whitespace between @samp{##} and its operands as you like, including
1536 comments, and you can put comments in arguments that will be
1537 concatenated. However, it is an error if @samp{##} appears at either
1538 end of a macro body.
1540 Consider a C program that interprets named commands. There probably
1541 needs to be a table of commands, perhaps an array of structures declared
1549 void (*function) (void);
1554 struct command commands[] =
1556 @{ "quit", quit_command @},
1557 @{ "help", help_command @},
1563 It would be cleaner not to have to give each command name twice, once in
1564 the string constant and once in the function name. A macro which takes the
1565 name of a command as an argument can make this unnecessary. The string
1566 constant can be created with stringification, and the function name by
1567 concatenating the argument with @samp{_command}. Here is how it is done:
1570 #define COMMAND(NAME) @{ #NAME, NAME ## _command @}
1572 struct command commands[] =
1580 @node Variadic Macros
1581 @section Variadic Macros
1582 @cindex variable number of arguments
1583 @cindex macros with variable arguments
1584 @cindex variadic macros
1586 A macro can be declared to accept a variable number of arguments much as
1587 a function can. The syntax for defining the macro is similar to that of
1588 a function. Here is an example:
1591 #define eprintf(@dots{}) fprintf (stderr, __VA_ARGS__)
1594 This kind of macro is called @dfn{variadic}. When the macro is invoked,
1595 all the tokens in its argument list after the last named argument (this
1596 macro has none), including any commas, become the @dfn{variable
1597 argument}. This sequence of tokens replaces the identifier
1598 @code{@w{__VA_ARGS__}} in the macro body wherever it appears. Thus, we
1599 have this expansion:
1602 eprintf ("%s:%d: ", input_file, lineno)
1603 @expansion{} fprintf (stderr, "%s:%d: ", input_file, lineno)
1606 The variable argument is completely macro-expanded before it is inserted
1607 into the macro expansion, just like an ordinary argument. You may use
1608 the @samp{#} and @samp{##} operators to stringify the variable argument
1609 or to paste its leading or trailing token with another token. (But see
1610 below for an important special case for @samp{##}.)
1612 If your macro is complicated, you may want a more descriptive name for
1613 the variable argument than @code{@w{__VA_ARGS__}}. CPP permits
1614 this, as an extension. You may write an argument name immediately
1615 before the @samp{@dots{}}; that name is used for the variable argument.
1616 The @code{eprintf} macro above could be written
1619 #define eprintf(args@dots{}) fprintf (stderr, args)
1623 using this extension. You cannot use @code{@w{__VA_ARGS__}} and this
1624 extension in the same macro.
1626 You can have named arguments as well as variable arguments in a variadic
1627 macro. We could define @code{eprintf} like this, instead:
1630 #define eprintf(format, @dots{}) fprintf (stderr, format, __VA_ARGS__)
1634 This formulation looks more descriptive, but unfortunately it is less
1635 flexible: you must now supply at least one argument after the format
1636 string. In standard C, you cannot omit the comma separating the named
1637 argument from the variable arguments. Furthermore, if you leave the
1638 variable argument empty, you will get a syntax error, because
1639 there will be an extra comma after the format string.
1642 eprintf("success!\n", );
1643 @expansion{} fprintf(stderr, "success!\n", );
1646 GNU CPP has a pair of extensions which deal with this problem. First,
1647 you are allowed to leave the variable argument out entirely:
1650 eprintf ("success!\n")
1651 @expansion{} fprintf(stderr, "success!\n", );
1655 Second, the @samp{##} token paste operator has a special meaning when
1656 placed between a comma and a variable argument. If you write
1659 #define eprintf(format, @dots{}) fprintf (stderr, format, ##__VA_ARGS__)
1663 and the variable argument is left out when the @code{eprintf} macro is
1664 used, then the comma before the @samp{##} will be deleted. This does
1665 @emph{not} happen if you pass an empty argument, nor does it happen if
1666 the token preceding @samp{##} is anything other than a comma.
1669 eprintf ("success!\n")
1670 @expansion{} fprintf(stderr, "success!\n");
1674 The above explanation is ambiguous about the case where the only macro
1675 parameter is a variable arguments parameter, as it is meaningless to
1676 try to distinguish whether no argument at all is an empty argument or
1677 a missing argument. In this case the C99 standard is clear that the
1678 comma must remain, however the existing GCC extension used to swallow
1679 the comma. So CPP retains the comma when conforming to a specific C
1680 standard, and drops it otherwise.
1682 C99 mandates that the only place the identifier @code{@w{__VA_ARGS__}}
1683 can appear is in the replacement list of a variadic macro. It may not
1684 be used as a macro name, macro argument name, or within a different type
1685 of macro. It may also be forbidden in open text; the standard is
1686 ambiguous. We recommend you avoid using it except for its defined
1689 Variadic macros are a new feature in C99. GNU CPP has supported them
1690 for a long time, but only with a named variable argument
1691 (@samp{args@dots{}}, not @samp{@dots{}} and @code{@w{__VA_ARGS__}}). If you are
1692 concerned with portability to previous versions of GCC, you should use
1693 only named variable arguments. On the other hand, if you are concerned
1694 with portability to other conforming implementations of C99, you should
1695 use only @code{@w{__VA_ARGS__}}.
1697 Previous versions of CPP implemented the comma-deletion extension
1698 much more generally. We have restricted it in this release to minimize
1699 the differences from C99. To get the same effect with both this and
1700 previous versions of GCC, the token preceding the special @samp{##} must
1701 be a comma, and there must be white space between that comma and
1702 whatever comes immediately before it:
1705 #define eprintf(format, args@dots{}) fprintf (stderr, format , ##args)
1709 @xref{Differences from previous versions}, for the gory details.
1711 @node Predefined Macros
1712 @section Predefined Macros
1714 @cindex predefined macros
1715 Several object-like macros are predefined; you use them without
1716 supplying their definitions. They fall into three classes: standard,
1717 common, and system-specific.
1719 In C++, there is a fourth category, the named operators. They act like
1720 predefined macros, but you cannot undefine them.
1723 * Standard Predefined Macros::
1724 * Common Predefined Macros::
1725 * System-specific Predefined Macros::
1726 * C++ Named Operators::
1729 @node Standard Predefined Macros
1730 @subsection Standard Predefined Macros
1731 @cindex standard predefined macros.
1733 The standard predefined macros are specified by the C and/or C++
1734 language standards, so they are available with all compilers that
1735 implement those standards. Older compilers may not provide all of
1736 them. Their names all start with double underscores.
1740 This macro expands to the name of the current input file, in the form of
1741 a C string constant. This is the path by which the preprocessor opened
1742 the file, not the short name specified in @samp{#include} or as the
1743 input file name argument. For example,
1744 @code{"/usr/local/include/myheader.h"} is a possible expansion of this
1748 This macro expands to the current input line number, in the form of a
1749 decimal integer constant. While we call it a predefined macro, it's
1750 a pretty strange macro, since its ``definition'' changes with each
1751 new line of source code.
1754 @code{__FILE__} and @code{__LINE__} are useful in generating an error
1755 message to report an inconsistency detected by the program; the message
1756 can state the source line at which the inconsistency was detected. For
1760 fprintf (stderr, "Internal error: "
1761 "negative string length "
1762 "%d at %s, line %d.",
1763 length, __FILE__, __LINE__);
1766 An @samp{#include} directive changes the expansions of @code{__FILE__}
1767 and @code{__LINE__} to correspond to the included file. At the end of
1768 that file, when processing resumes on the input file that contained
1769 the @samp{#include} directive, the expansions of @code{__FILE__} and
1770 @code{__LINE__} revert to the values they had before the
1771 @samp{#include} (but @code{__LINE__} is then incremented by one as
1772 processing moves to the line after the @samp{#include}).
1774 A @samp{#line} directive changes @code{__LINE__}, and may change
1775 @code{__FILE__} as well. @xref{Line Control}.
1777 C99 introduces @code{__func__}, and GCC has provided @code{__FUNCTION__}
1778 for a long time. Both of these are strings containing the name of the
1779 current function (there are slight semantic differences; see the GCC
1780 manual). Neither of them is a macro; the preprocessor does not know the
1781 name of the current function. They tend to be useful in conjunction
1782 with @code{__FILE__} and @code{__LINE__}, though.
1787 This macro expands to a string constant that describes the date on which
1788 the preprocessor is being run. The string constant contains eleven
1789 characters and looks like @code{@w{"Feb 12 1996"}}. If the day of the
1790 month is less than 10, it is padded with a space on the left.
1792 If GCC cannot determine the current date, it will emit a warning message
1793 (once per compilation) and @code{__DATE__} will expand to
1794 @code{@w{"??? ?? ????"}}.
1797 This macro expands to a string constant that describes the time at
1798 which the preprocessor is being run. The string constant contains
1799 eight characters and looks like @code{"23:59:01"}.
1801 If GCC cannot determine the current time, it will emit a warning message
1802 (once per compilation) and @code{__TIME__} will expand to
1806 In normal operation, this macro expands to the constant 1, to signify
1807 that this compiler conforms to ISO Standard C@. If GNU CPP is used with
1808 a compiler other than GCC, this is not necessarily true; however, the
1809 preprocessor always conforms to the standard unless the
1810 @option{-traditional-cpp} option is used.
1812 This macro is not defined if the @option{-traditional-cpp} option is used.
1814 On some hosts, the system compiler uses a different convention, where
1815 @code{__STDC__} is normally 0, but is 1 if the user specifies strict
1816 conformance to the C Standard. CPP follows the host convention when
1817 processing system header files, but when processing user files
1818 @code{__STDC__} is always 1. This has been reported to cause problems;
1819 for instance, some versions of Solaris provide X Windows headers that
1820 expect @code{__STDC__} to be either undefined or 1. You may be able to
1821 work around this sort of problem by using an @option{-I} option to
1822 cancel treatment of those headers as system headers. @xref{Invocation}.
1824 @item __STDC_VERSION__
1825 This macro expands to the C Standard's version number, a long integer
1826 constant of the form @code{@var{yyyy}@var{mm}L} where @var{yyyy} and
1827 @var{mm} are the year and month of the Standard version. This signifies
1828 which version of the C Standard the compiler conforms to. Like
1829 @code{__STDC__}, this is not necessarily accurate for the entire
1830 implementation, unless GNU CPP is being used with GCC@.
1832 The value @code{199409L} signifies the 1989 C standard as amended in
1833 1994, which is the current default; the value @code{199901L} signifies
1834 the 1999 revision of the C standard. Support for the 1999 revision is
1837 This macro is not defined if the @option{-traditional-cpp} option is
1838 used, nor when compiling C++ or Objective-C@.
1840 @item __STDC_HOSTED__
1841 This macro is defined, with value 1, if the compiler's target is a
1842 @dfn{hosted environment}. A hosted environment has the complete
1843 facilities of the standard C library available.
1846 This macro is defined when the C++ compiler is in use. You can use
1847 @code{__cplusplus} to test whether a header is compiled by a C compiler
1848 or a C++ compiler. This macro is similar to @code{__STDC_VERSION__}, in
1849 that it expands to a version number. A fully conforming implementation
1850 of the 1998 C++ standard will define this macro to @code{199711L}. The
1851 GNU C++ compiler is not yet fully conforming, so it uses @code{1}
1852 instead. We hope to complete our implementation in the near future.
1856 @node Common Predefined Macros
1857 @subsection Common Predefined Macros
1858 @cindex common predefined macros
1860 The common predefined macros are GNU C extensions. They are available
1861 with the same meanings regardless of the machine or operating system on
1862 which you are using GNU C@. Their names all start with double
1868 @itemx __GNUC_MINOR__
1869 @itemx __GNUC_PATCHLEVEL__
1870 These macros are defined by all GNU compilers that use the C
1871 preprocessor: C, C++, and Objective-C@. Their values are the major
1872 version, minor version, and patch level of the compiler, as integer
1873 constants. For example, GCC 3.2.1 will define @code{__GNUC__} to 3,
1874 @code{__GNUC_MINOR__} to 2, and @code{__GNUC_PATCHLEVEL__} to 1. They
1875 are defined only when the entire compiler is in use; if you invoke the
1876 preprocessor directly, they are not defined.
1878 @code{__GNUC_PATCHLEVEL__} is new to GCC 3.0; it is also present in the
1879 widely-used development snapshots leading up to 3.0 (which identify
1880 themselves as GCC 2.96 or 2.97, depending on which snapshot you have).
1882 If all you need to know is whether or not your program is being compiled
1883 by GCC, you can simply test @code{__GNUC__}. If you need to write code
1884 which depends on a specific version, you must be more careful. Each
1885 time the minor version is increased, the patch level is reset to zero;
1886 each time the major version is increased (which happens rarely), the
1887 minor version and patch level are reset. If you wish to use the
1888 predefined macros directly in the conditional, you will need to write it
1892 /* @r{Test for GCC > 3.2.0} */
1893 #if __GNUC__ > 3 || \
1894 (__GNUC__ == 3 && (__GNUC_MINOR__ > 2 || \
1895 (__GNUC_MINOR__ == 2 && \
1896 __GNUC_PATCHLEVEL__ > 0))
1900 Another approach is to use the predefined macros to
1901 calculate a single number, then compare that against a threshold:
1904 #define GCC_VERSION (__GNUC__ * 10000 \
1905 + __GNUC_MINOR__ * 100 \
1906 + __GNUC_PATCHLEVEL__)
1908 /* @r{Test for GCC > 3.2.0} */
1909 #if GCC_VERSION > 30200
1913 Many people find this form easier to understand.
1916 This macro is defined, with value 1, when the Objective-C compiler is in
1917 use. You can use @code{__OBJC__} to test whether a header is compiled
1918 by a C compiler or a Objective-C compiler.
1921 The GNU C++ compiler defines this. Testing it is equivalent to
1922 testing @code{@w{(__GNUC__ && __cplusplus)}}.
1924 @item __STRICT_ANSI__
1925 GCC defines this macro if and only if the @option{-ansi} switch, or a
1926 @option{-std} switch specifying strict conformance to some version of ISO C,
1927 was specified when GCC was invoked. It is defined to @samp{1}.
1928 This macro exists primarily to direct GNU libc's header files to
1929 restrict their definitions to the minimal set found in the 1989 C
1933 This macro expands to the name of the main input file, in the form
1934 of a C string constant. This is the source file that was specified
1935 on the command line of the preprocessor or C compiler.
1937 @item __INCLUDE_LEVEL__
1938 This macro expands to a decimal integer constant that represents the
1939 depth of nesting in include files. The value of this macro is
1940 incremented on every @samp{#include} directive and decremented at the
1941 end of every included file. It starts out at 0, it's value within the
1942 base file specified on the command line.
1945 This macro expands to a string constant which describes the version of
1946 the compiler in use. You should not rely on its contents having any
1947 particular form, but it can be counted on to contain at least the
1951 @itemx __OPTIMIZE_SIZE__
1952 @itemx __NO_INLINE__
1953 These macros describe the compilation mode. @code{__OPTIMIZE__} is
1954 defined in all optimizing compilations. @code{__OPTIMIZE_SIZE__} is
1955 defined if the compiler is optimizing for size, not speed.
1956 @code{__NO_INLINE__} is defined if no functions will be inlined into
1957 their callers (when not optimizing, or when inlining has been
1958 specifically disabled by @option{-fno-inline}).
1960 These macros cause certain GNU header files to provide optimized
1961 definitions, using macros or inline functions, of system library
1962 functions. You should not use these macros in any way unless you make
1963 sure that programs will execute with the same effect whether or not they
1964 are defined. If they are defined, their value is 1.
1966 @item __CHAR_UNSIGNED__
1967 GCC defines this macro if and only if the data type @code{char} is
1968 unsigned on the target machine. It exists to cause the standard header
1969 file @file{limits.h} to work correctly. You should not use this macro
1970 yourself; instead, refer to the standard macros defined in @file{limits.h}.
1972 @item __REGISTER_PREFIX__
1973 This macro expands to a single token (not a string constant) which is
1974 the prefix applied to CPU register names in assembly language for this
1975 target. You can use it to write assembly that is usable in multiple
1976 environments. For example, in the @code{m68k-aout} environment it
1977 expands to nothing, but in the @code{m68k-coff} environment it expands
1978 to a single @samp{%}.
1980 @item __USER_LABEL_PREFIX__
1981 This macro expands to a single token which is the prefix applied to
1982 user labels (symbols visible to C code) in assembly. For example, in
1983 the @code{m68k-aout} environment it expands to an @samp{_}, but in the
1984 @code{m68k-coff} environment it expands to nothing.
1986 This macro will have the correct definition even if
1987 @option{-f(no-)underscores} is in use, but it will not be correct if
1988 target-specific options that adjust this prefix are used (e.g.@: the
1989 OSF/rose @option{-mno-underscores} option).
1992 @itemx __PTRDIFF_TYPE__
1993 @itemx __WCHAR_TYPE__
1994 @itemx __WINT_TYPE__
1995 These macros are defined to the correct underlying types for the
1996 @code{size_t}, @code{ptrdiff_t}, @code{wchar_t}, and @code{wint_t}
1997 typedefs, respectively. They exist to make the standard header files
1998 @file{stddef.h} and @file{wchar.h} work correctly. You should not use
1999 these macros directly; instead, include the appropriate headers and use
2002 @item __USING_SJLJ_EXCEPTIONS__
2003 This macro is defined, with value 1, if the compiler uses the old
2004 mechanism based on @code{setjmp} and @code{longjmp} for exception
2008 @node System-specific Predefined Macros
2009 @subsection System-specific Predefined Macros
2011 @cindex system-specific predefined macros
2012 @cindex predefined macros, system-specific
2013 @cindex reserved namespace
2015 The C preprocessor normally predefines several macros that indicate what
2016 type of system and machine is in use. They are obviously different on
2017 each target supported by GCC@. This manual, being for all systems and
2018 machines, cannot tell you what their names are, but you can use
2019 @command{cpp -dM} to see them all. @xref{Invocation}. All system-specific
2020 predefined macros expand to the constant 1, so you can test them with
2021 either @samp{#ifdef} or @samp{#if}.
2023 The C standard requires that all system-specific macros be part of the
2024 @dfn{reserved namespace}. All names which begin with two underscores,
2025 or an underscore and a capital letter, are reserved for the compiler and
2026 library to use as they wish. However, historically system-specific
2027 macros have had names with no special prefix; for instance, it is common
2028 to find @code{unix} defined on Unix systems. For all such macros, GCC
2029 provides a parallel macro with two underscores added at the beginning
2030 and the end. If @code{unix} is defined, @code{__unix__} will be defined
2031 too. There will never be more than two underscores; the parallel of
2032 @code{_mips} is @code{__mips__}.
2034 When the @option{-ansi} option, or any @option{-std} option that
2035 requests strict conformance, is given to the compiler, all the
2036 system-specific predefined macros outside the reserved namespace are
2037 suppressed. The parallel macros, inside the reserved namespace, remain
2040 We are slowly phasing out all predefined macros which are outside the
2041 reserved namespace. You should never use them in new programs, and we
2042 encourage you to correct older code to use the parallel macros whenever
2043 you find it. We don't recommend you use the system-specific macros that
2044 are in the reserved namespace, either. It is better in the long run to
2045 check specifically for features you need, using a tool such as
2048 @node C++ Named Operators
2049 @subsection C++ Named Operators
2050 @cindex named operators
2051 @cindex C++ named operators
2054 In C++, there are eleven keywords which are simply alternate spellings
2055 of operators normally written with punctuation. These keywords are
2056 treated as such even in the preprocessor. They function as operators in
2057 @samp{#if}, and they cannot be defined as macros or poisoned. In C, you
2058 can request that those keywords take their C++ meaning by including
2059 @file{iso646.h}. That header defines each one as a normal object-like
2060 macro expanding to the appropriate punctuator.
2062 These are the named operators and their corresponding punctuators:
2064 @multitable {Named Operator} {Punctuator}
2065 @item Named Operator @tab Punctuator
2066 @item @code{and} @tab @code{&&}
2067 @item @code{and_eq} @tab @code{&=}
2068 @item @code{bitand} @tab @code{&}
2069 @item @code{bitor} @tab @code{|}
2070 @item @code{compl} @tab @code{~}
2071 @item @code{not} @tab @code{!}
2072 @item @code{not_eq} @tab @code{!=}
2073 @item @code{or} @tab @code{||}
2074 @item @code{or_eq} @tab @code{|=}
2075 @item @code{xor} @tab @code{^}
2076 @item @code{xor_eq} @tab @code{^=}
2079 @node Undefining and Redefining Macros
2080 @section Undefining and Redefining Macros
2081 @cindex undefining macros
2082 @cindex redefining macros
2085 If a macro ceases to be useful, it may be @dfn{undefined} with the
2086 @samp{#undef} directive. @samp{#undef} takes a single argument, the
2087 name of the macro to undefine. You use the bare macro name, even if the
2088 macro is function-like. It is an error if anything appears on the line
2089 after the macro name. @samp{#undef} has no effect if the name is not a
2094 x = FOO; @expansion{} x = 4;
2096 x = FOO; @expansion{} x = FOO;
2099 Once a macro has been undefined, that identifier may be @dfn{redefined}
2100 as a macro by a subsequent @samp{#define} directive. The new definition
2101 need not have any resemblance to the old definition.
2103 However, if an identifier which is currently a macro is redefined, then
2104 the new definition must be @dfn{effectively the same} as the old one.
2105 Two macro definitions are effectively the same if:
2107 @item Both are the same type of macro (object- or function-like).
2108 @item All the tokens of the replacement list are the same.
2109 @item If there are any parameters, they are the same.
2110 @item Whitespace appears in the same places in both. It need not be
2111 exactly the same amount of whitespace, though. Remember that comments
2112 count as whitespace.
2116 These definitions are effectively the same:
2118 #define FOUR (2 + 2)
2119 #define FOUR (2 + 2)
2120 #define FOUR (2 /* two */ + 2)
2125 #define FOUR (2 + 2)
2126 #define FOUR ( 2+2 )
2127 #define FOUR (2 * 2)
2128 #define FOUR(score,and,seven,years,ago) (2 + 2)
2131 If a macro is redefined with a definition that is not effectively the
2132 same as the old one, the preprocessor issues a warning and changes the
2133 macro to use the new definition. If the new definition is effectively
2134 the same, the redefinition is silently ignored. This allows, for
2135 instance, two different headers to define a common macro. The
2136 preprocessor will only complain if the definitions do not match.
2138 @node Directives Within Macro Arguments
2139 @section Directives Within Macro Arguments
2140 @cindex macro arguments and directives
2142 Occasionally it is convenient to use preprocessor directives within
2143 the arguments of a macro. The C and C++ standards declare that
2144 behavior in these cases is undefined.
2146 Versions of CPP prior to 3.2 would reject such constructs with an
2147 error message. This was the only syntactic difference between normal
2148 functions and function-like macros, so it seemed attractive to remove
2149 this limitation, and people would often be surprised that they could
2150 not use macros in this way. Moreover, sometimes people would use
2151 conditional compilation in the argument list to a normal library
2152 function like @samp{printf}, only to find that after a library upgrade
2153 @samp{printf} had changed to be a function-like macro, and their code
2154 would no longer compile. So from version 3.2 we changed CPP to
2155 successfully process arbitrary directives within macro arguments in
2156 exactly the same way as it would have processed the directive were the
2157 function-like macro invocation not present.
2159 If, within a macro invocation, that macro is redefined, then the new
2160 definition takes effect in time for argument pre-expansion, but the
2161 original definition is still used for argument replacement. Here is a
2162 pathological example:
2180 with the semantics described above.
2182 @node Macro Pitfalls
2183 @section Macro Pitfalls
2184 @cindex problems with macros
2185 @cindex pitfalls of macros
2187 In this section we describe some special rules that apply to macros and
2188 macro expansion, and point out certain cases in which the rules have
2189 counter-intuitive consequences that you must watch out for.
2193 * Operator Precedence Problems::
2194 * Swallowing the Semicolon::
2195 * Duplication of Side Effects::
2196 * Self-Referential Macros::
2197 * Argument Prescan::
2198 * Newlines in Arguments::
2202 @subsection Misnesting
2204 When a macro is called with arguments, the arguments are substituted
2205 into the macro body and the result is checked, together with the rest of
2206 the input file, for more macro calls. It is possible to piece together
2207 a macro call coming partially from the macro body and partially from the
2208 arguments. For example,
2211 #define twice(x) (2*(x))
2212 #define call_with_1(x) x(1)
2214 @expansion{} twice(1)
2215 @expansion{} (2*(1))
2218 Macro definitions do not have to have balanced parentheses. By writing
2219 an unbalanced open parenthesis in a macro body, it is possible to create
2220 a macro call that begins inside the macro body but ends outside of it.
2224 #define strange(file) fprintf (file, "%s %d",
2226 strange(stderr) p, 35)
2227 @expansion{} fprintf (stderr, "%s %d", p, 35)
2230 The ability to piece together a macro call can be useful, but the use of
2231 unbalanced open parentheses in a macro body is just confusing, and
2234 @node Operator Precedence Problems
2235 @subsection Operator Precedence Problems
2236 @cindex parentheses in macro bodies
2238 You may have noticed that in most of the macro definition examples shown
2239 above, each occurrence of a macro argument name had parentheses around
2240 it. In addition, another pair of parentheses usually surround the
2241 entire macro definition. Here is why it is best to write macros that
2244 Suppose you define a macro as follows,
2247 #define ceil_div(x, y) (x + y - 1) / y
2251 whose purpose is to divide, rounding up. (One use for this operation is
2252 to compute how many @code{int} objects are needed to hold a certain
2253 number of @code{char} objects.) Then suppose it is used as follows:
2256 a = ceil_div (b & c, sizeof (int));
2257 @expansion{} a = (b & c + sizeof (int) - 1) / sizeof (int);
2261 This does not do what is intended. The operator-precedence rules of
2262 C make it equivalent to this:
2265 a = (b & (c + sizeof (int) - 1)) / sizeof (int);
2269 What we want is this:
2272 a = ((b & c) + sizeof (int) - 1)) / sizeof (int);
2276 Defining the macro as
2279 #define ceil_div(x, y) ((x) + (y) - 1) / (y)
2283 provides the desired result.
2285 Unintended grouping can result in another way. Consider @code{sizeof
2286 ceil_div(1, 2)}. That has the appearance of a C expression that would
2287 compute the size of the type of @code{ceil_div (1, 2)}, but in fact it
2288 means something very different. Here is what it expands to:
2291 sizeof ((1) + (2) - 1) / (2)
2295 This would take the size of an integer and divide it by two. The
2296 precedence rules have put the division outside the @code{sizeof} when it
2297 was intended to be inside.
2299 Parentheses around the entire macro definition prevent such problems.
2300 Here, then, is the recommended way to define @code{ceil_div}:
2303 #define ceil_div(x, y) (((x) + (y) - 1) / (y))
2306 @node Swallowing the Semicolon
2307 @subsection Swallowing the Semicolon
2308 @cindex semicolons (after macro calls)
2310 Often it is desirable to define a macro that expands into a compound
2311 statement. Consider, for example, the following macro, that advances a
2312 pointer (the argument @code{p} says where to find it) across whitespace
2316 #define SKIP_SPACES(p, limit) \
2317 @{ char *lim = (limit); \
2318 while (p < lim) @{ \
2319 if (*p++ != ' ') @{ \
2324 Here backslash-newline is used to split the macro definition, which must
2325 be a single logical line, so that it resembles the way such code would
2326 be laid out if not part of a macro definition.
2328 A call to this macro might be @code{SKIP_SPACES (p, lim)}. Strictly
2329 speaking, the call expands to a compound statement, which is a complete
2330 statement with no need for a semicolon to end it. However, since it
2331 looks like a function call, it minimizes confusion if you can use it
2332 like a function call, writing a semicolon afterward, as in
2333 @code{SKIP_SPACES (p, lim);}
2335 This can cause trouble before @code{else} statements, because the
2336 semicolon is actually a null statement. Suppose you write
2340 SKIP_SPACES (p, lim);
2345 The presence of two statements---the compound statement and a null
2346 statement---in between the @code{if} condition and the @code{else}
2347 makes invalid C code.
2349 The definition of the macro @code{SKIP_SPACES} can be altered to solve
2350 this problem, using a @code{do @dots{} while} statement. Here is how:
2353 #define SKIP_SPACES(p, limit) \
2354 do @{ char *lim = (limit); \
2355 while (p < lim) @{ \
2356 if (*p++ != ' ') @{ \
2357 p--; break; @}@}@} \
2361 Now @code{SKIP_SPACES (p, lim);} expands into
2364 do @{@dots{}@} while (0);
2368 which is one statement. The loop executes exactly once; most compilers
2369 generate no extra code for it.
2371 @node Duplication of Side Effects
2372 @subsection Duplication of Side Effects
2374 @cindex side effects (in macro arguments)
2375 @cindex unsafe macros
2376 Many C programs define a macro @code{min}, for ``minimum'', like this:
2379 #define min(X, Y) ((X) < (Y) ? (X) : (Y))
2382 When you use this macro with an argument containing a side effect,
2386 next = min (x + y, foo (z));
2390 it expands as follows:
2393 next = ((x + y) < (foo (z)) ? (x + y) : (foo (z)));
2397 where @code{x + y} has been substituted for @code{X} and @code{foo (z)}
2400 The function @code{foo} is used only once in the statement as it appears
2401 in the program, but the expression @code{foo (z)} has been substituted
2402 twice into the macro expansion. As a result, @code{foo} might be called
2403 two times when the statement is executed. If it has side effects or if
2404 it takes a long time to compute, the results might not be what you
2405 intended. We say that @code{min} is an @dfn{unsafe} macro.
2407 The best solution to this problem is to define @code{min} in a way that
2408 computes the value of @code{foo (z)} only once. The C language offers
2409 no standard way to do this, but it can be done with GNU extensions as
2414 (@{ typeof (X) x_ = (X); \
2415 typeof (Y) y_ = (Y); \
2416 (x_ < y_) ? x_ : y_; @})
2419 The @samp{(@{ @dots{} @})} notation produces a compound statement that
2420 acts as an expression. Its value is the value of its last statement.
2421 This permits us to define local variables and assign each argument to
2422 one. The local variables have underscores after their names to reduce
2423 the risk of conflict with an identifier of wider scope (it is impossible
2424 to avoid this entirely). Now each argument is evaluated exactly once.
2426 If you do not wish to use GNU C extensions, the only solution is to be
2427 careful when @emph{using} the macro @code{min}. For example, you can
2428 calculate the value of @code{foo (z)}, save it in a variable, and use
2429 that variable in @code{min}:
2433 #define min(X, Y) ((X) < (Y) ? (X) : (Y))
2437 next = min (x + y, tem);
2443 (where we assume that @code{foo} returns type @code{int}).
2445 @node Self-Referential Macros
2446 @subsection Self-Referential Macros
2447 @cindex self-reference
2449 A @dfn{self-referential} macro is one whose name appears in its
2450 definition. Recall that all macro definitions are rescanned for more
2451 macros to replace. If the self-reference were considered a use of the
2452 macro, it would produce an infinitely large expansion. To prevent this,
2453 the self-reference is not considered a macro call. It is passed into
2454 the preprocessor output unchanged. Let's consider an example:
2457 #define foo (4 + foo)
2461 where @code{foo} is also a variable in your program.
2463 Following the ordinary rules, each reference to @code{foo} will expand
2464 into @code{(4 + foo)}; then this will be rescanned and will expand into
2465 @code{(4 + (4 + foo))}; and so on until the computer runs out of memory.
2467 The self-reference rule cuts this process short after one step, at
2468 @code{(4 + foo)}. Therefore, this macro definition has the possibly
2469 useful effect of causing the program to add 4 to the value of @code{foo}
2470 wherever @code{foo} is referred to.
2472 In most cases, it is a bad idea to take advantage of this feature. A
2473 person reading the program who sees that @code{foo} is a variable will
2474 not expect that it is a macro as well. The reader will come across the
2475 identifier @code{foo} in the program and think its value should be that
2476 of the variable @code{foo}, whereas in fact the value is four greater.
2478 One common, useful use of self-reference is to create a macro which
2479 expands to itself. If you write
2486 then the macro @code{EPERM} expands to @code{EPERM}. Effectively, it is
2487 left alone by the preprocessor whenever it's used in running text. You
2488 can tell that it's a macro with @samp{#ifdef}. You might do this if you
2489 want to define numeric constants with an @code{enum}, but have
2490 @samp{#ifdef} be true for each constant.
2492 If a macro @code{x} expands to use a macro @code{y}, and the expansion of
2493 @code{y} refers to the macro @code{x}, that is an @dfn{indirect
2494 self-reference} of @code{x}. @code{x} is not expanded in this case
2495 either. Thus, if we have
2503 then @code{x} and @code{y} expand as follows:
2507 x @expansion{} (4 + y)
2508 @expansion{} (4 + (2 * x))
2510 y @expansion{} (2 * x)
2511 @expansion{} (2 * (4 + y))
2516 Each macro is expanded when it appears in the definition of the other
2517 macro, but not when it indirectly appears in its own definition.
2519 @node Argument Prescan
2520 @subsection Argument Prescan
2521 @cindex expansion of arguments
2522 @cindex macro argument expansion
2523 @cindex prescan of macro arguments
2525 Macro arguments are completely macro-expanded before they are
2526 substituted into a macro body, unless they are stringified or pasted
2527 with other tokens. After substitution, the entire macro body, including
2528 the substituted arguments, is scanned again for macros to be expanded.
2529 The result is that the arguments are scanned @emph{twice} to expand
2530 macro calls in them.
2532 Most of the time, this has no effect. If the argument contained any
2533 macro calls, they are expanded during the first scan. The result
2534 therefore contains no macro calls, so the second scan does not change
2535 it. If the argument were substituted as given, with no prescan, the
2536 single remaining scan would find the same macro calls and produce the
2539 You might expect the double scan to change the results when a
2540 self-referential macro is used in an argument of another macro
2541 (@pxref{Self-Referential Macros}): the self-referential macro would be
2542 expanded once in the first scan, and a second time in the second scan.
2543 However, this is not what happens. The self-references that do not
2544 expand in the first scan are marked so that they will not expand in the
2547 You might wonder, ``Why mention the prescan, if it makes no difference?
2548 And why not skip it and make the preprocessor faster?'' The answer is
2549 that the prescan does make a difference in three special cases:
2553 Nested calls to a macro.
2555 We say that @dfn{nested} calls to a macro occur when a macro's argument
2556 contains a call to that very macro. For example, if @code{f} is a macro
2557 that expects one argument, @code{f (f (1))} is a nested pair of calls to
2558 @code{f}. The desired expansion is made by expanding @code{f (1)} and
2559 substituting that into the definition of @code{f}. The prescan causes
2560 the expected result to happen. Without the prescan, @code{f (1)} itself
2561 would be substituted as an argument, and the inner use of @code{f} would
2562 appear during the main scan as an indirect self-reference and would not
2566 Macros that call other macros that stringify or concatenate.
2568 If an argument is stringified or concatenated, the prescan does not
2569 occur. If you @emph{want} to expand a macro, then stringify or
2570 concatenate its expansion, you can do that by causing one macro to call
2571 another macro that does the stringification or concatenation. For
2572 instance, if you have
2575 #define AFTERX(x) X_ ## x
2576 #define XAFTERX(x) AFTERX(x)
2577 #define TABLESIZE 1024
2578 #define BUFSIZE TABLESIZE
2581 then @code{AFTERX(BUFSIZE)} expands to @code{X_BUFSIZE}, and
2582 @code{XAFTERX(BUFSIZE)} expands to @code{X_1024}. (Not to
2583 @code{X_TABLESIZE}. Prescan always does a complete expansion.)
2586 Macros used in arguments, whose expansions contain unshielded commas.
2588 This can cause a macro expanded on the second scan to be called with the
2589 wrong number of arguments. Here is an example:
2593 #define bar(x) lose(x)
2594 #define lose(x) (1 + (x))
2597 We would like @code{bar(foo)} to turn into @code{(1 + (foo))}, which
2598 would then turn into @code{(1 + (a,b))}. Instead, @code{bar(foo)}
2599 expands into @code{lose(a,b)}, and you get an error because @code{lose}
2600 requires a single argument. In this case, the problem is easily solved
2601 by the same parentheses that ought to be used to prevent misnesting of
2602 arithmetic operations:
2607 #define bar(x) lose((x))
2610 The extra pair of parentheses prevents the comma in @code{foo}'s
2611 definition from being interpreted as an argument separator.
2615 @node Newlines in Arguments
2616 @subsection Newlines in Arguments
2617 @cindex newlines in macro arguments
2619 The invocation of a function-like macro can extend over many logical
2620 lines. However, in the present implementation, the entire expansion
2621 comes out on one line. Thus line numbers emitted by the compiler or
2622 debugger refer to the line the invocation started on, which might be
2623 different to the line containing the argument causing the problem.
2625 Here is an example illustrating this:
2628 #define ignore_second_arg(a,b,c) a; c
2630 ignore_second_arg (foo (),
2636 The syntax error triggered by the tokens @code{syntax error} results in
2637 an error message citing line three---the line of ignore_second_arg---
2638 even though the problematic code comes from line five.
2640 We consider this a bug, and intend to fix it in the near future.
2643 @chapter Conditionals
2644 @cindex conditionals
2646 A @dfn{conditional} is a directive that instructs the preprocessor to
2647 select whether or not to include a chunk of code in the final token
2648 stream passed to the compiler. Preprocessor conditionals can test
2649 arithmetic expressions, or whether a name is defined as a macro, or both
2650 simultaneously using the special @code{defined} operator.
2652 A conditional in the C preprocessor resembles in some ways an @code{if}
2653 statement in C, but it is important to understand the difference between
2654 them. The condition in an @code{if} statement is tested during the
2655 execution of your program. Its purpose is to allow your program to
2656 behave differently from run to run, depending on the data it is
2657 operating on. The condition in a preprocessing conditional directive is
2658 tested when your program is compiled. Its purpose is to allow different
2659 code to be included in the program depending on the situation at the
2660 time of compilation.
2662 However, the distinction is becoming less clear. Modern compilers often
2663 do test @code{if} statements when a program is compiled, if their
2664 conditions are known not to vary at run time, and eliminate code which
2665 can never be executed. If you can count on your compiler to do this,
2666 you may find that your program is more readable if you use @code{if}
2667 statements with constant conditions (perhaps determined by macros). Of
2668 course, you can only use this to exclude code, not type definitions or
2669 other preprocessing directives, and you can only do it if the code
2670 remains syntactically valid when it is not to be used.
2672 GCC version 3 eliminates this kind of never-executed code even when
2673 not optimizing. Older versions did it only when optimizing.
2676 * Conditional Uses::
2677 * Conditional Syntax::
2681 @node Conditional Uses
2682 @section Conditional Uses
2684 There are three general reasons to use a conditional.
2688 A program may need to use different code depending on the machine or
2689 operating system it is to run on. In some cases the code for one
2690 operating system may be erroneous on another operating system; for
2691 example, it might refer to data types or constants that do not exist on
2692 the other system. When this happens, it is not enough to avoid
2693 executing the invalid code. Its mere presence will cause the compiler
2694 to reject the program. With a preprocessing conditional, the offending
2695 code can be effectively excised from the program when it is not valid.
2698 You may want to be able to compile the same source file into two
2699 different programs. One version might make frequent time-consuming
2700 consistency checks on its intermediate data, or print the values of
2701 those data for debugging, and the other not.
2704 A conditional whose condition is always false is one way to exclude code
2705 from the program but keep it as a sort of comment for future reference.
2708 Simple programs that do not need system-specific logic or complex
2709 debugging hooks generally will not need to use preprocessing
2712 @node Conditional Syntax
2713 @section Conditional Syntax
2716 A conditional in the C preprocessor begins with a @dfn{conditional
2717 directive}: @samp{#if}, @samp{#ifdef} or @samp{#ifndef}.
2732 The simplest sort of conditional is
2738 @var{controlled text}
2740 #endif /* @var{MACRO} */
2744 @cindex conditional group
2745 This block is called a @dfn{conditional group}. @var{controlled text}
2746 will be included in the output of the preprocessor if and only if
2747 @var{MACRO} is defined. We say that the conditional @dfn{succeeds} if
2748 @var{MACRO} is defined, @dfn{fails} if it is not.
2750 The @var{controlled text} inside of a conditional can include
2751 preprocessing directives. They are executed only if the conditional
2752 succeeds. You can nest conditional groups inside other conditional
2753 groups, but they must be completely nested. In other words,
2754 @samp{#endif} always matches the nearest @samp{#ifdef} (or
2755 @samp{#ifndef}, or @samp{#if}). Also, you cannot start a conditional
2756 group in one file and end it in another.
2758 Even if a conditional fails, the @var{controlled text} inside it is
2759 still run through initial transformations and tokenization. Therefore,
2760 it must all be lexically valid C@. Normally the only way this matters is
2761 that all comments and string literals inside a failing conditional group
2762 must still be properly ended.
2764 The comment following the @samp{#endif} is not required, but it is a
2765 good practice if there is a lot of @var{controlled text}, because it
2766 helps people match the @samp{#endif} to the corresponding @samp{#ifdef}.
2767 Older programs sometimes put @var{MACRO} directly after the
2768 @samp{#endif} without enclosing it in a comment. This is invalid code
2769 according to the C standard. CPP accepts it with a warning. It
2770 never affects which @samp{#ifndef} the @samp{#endif} matches.
2773 Sometimes you wish to use some code if a macro is @emph{not} defined.
2774 You can do this by writing @samp{#ifndef} instead of @samp{#ifdef}.
2775 One common use of @samp{#ifndef} is to include code only the first
2776 time a header file is included. @xref{Once-Only Headers}.
2778 Macro definitions can vary between compilations for several reasons.
2779 Here are some samples.
2783 Some macros are predefined on each kind of machine
2784 (@pxref{System-specific Predefined Macros}). This allows you to provide
2785 code specially tuned for a particular machine.
2788 System header files define more macros, associated with the features
2789 they implement. You can test these macros with conditionals to avoid
2790 using a system feature on a machine where it is not implemented.
2793 Macros can be defined or undefined with the @option{-D} and @option{-U}
2794 command line options when you compile the program. You can arrange to
2795 compile the same source file into two different programs by choosing a
2796 macro name to specify which program you want, writing conditionals to
2797 test whether or how this macro is defined, and then controlling the
2798 state of the macro with command line options, perhaps set in the
2799 Makefile. @xref{Invocation}.
2802 Your program might have a special header file (often called
2803 @file{config.h}) that is adjusted when the program is compiled. It can
2804 define or not define macros depending on the features of the system and
2805 the desired capabilities of the program. The adjustment can be
2806 automated by a tool such as @command{autoconf}, or done by hand.
2812 The @samp{#if} directive allows you to test the value of an arithmetic
2813 expression, rather than the mere existence of one macro. Its syntax is
2817 #if @var{expression}
2819 @var{controlled text}
2821 #endif /* @var{expression} */
2825 @var{expression} is a C expression of integer type, subject to stringent
2826 restrictions. It may contain
2833 Character constants, which are interpreted as they would be in normal
2837 Arithmetic operators for addition, subtraction, multiplication,
2838 division, bitwise operations, shifts, comparisons, and logical
2839 operations (@code{&&} and @code{||}). The latter two obey the usual
2840 short-circuiting rules of standard C@.
2843 Macros. All macros in the expression are expanded before actual
2844 computation of the expression's value begins.
2847 Uses of the @code{defined} operator, which lets you check whether macros
2848 are defined in the middle of an @samp{#if}.
2851 Identifiers that are not macros, which are all considered to be the
2852 number zero. This allows you to write @code{@w{#if MACRO}} instead of
2853 @code{@w{#ifdef MACRO}}, if you know that MACRO, when defined, will
2854 always have a nonzero value. Function-like macros used without their
2855 function call parentheses are also treated as zero.
2857 In some contexts this shortcut is undesirable. The @option{-Wundef}
2858 option causes GCC to warn whenever it encounters an identifier which is
2859 not a macro in an @samp{#if}.
2862 The preprocessor does not know anything about types in the language.
2863 Therefore, @code{sizeof} operators are not recognized in @samp{#if}, and
2864 neither are @code{enum} constants. They will be taken as identifiers
2865 which are not macros, and replaced by zero. In the case of
2866 @code{sizeof}, this is likely to cause the expression to be invalid.
2868 The preprocessor calculates the value of @var{expression}. It carries
2869 out all calculations in the widest integer type known to the compiler;
2870 on most machines supported by GCC this is 64 bits. This is not the same
2871 rule as the compiler uses to calculate the value of a constant
2872 expression, and may give different results in some cases. If the value
2873 comes out to be nonzero, the @samp{#if} succeeds and the @var{controlled
2874 text} is included; otherwise it is skipped.
2876 If @var{expression} is not correctly formed, GCC issues an error and
2877 treats the conditional as having failed.
2882 @cindex @code{defined}
2883 The special operator @code{defined} is used in @samp{#if} and
2884 @samp{#elif} expressions to test whether a certain name is defined as a
2885 macro. @code{defined @var{name}} and @code{defined (@var{name})} are
2886 both expressions whose value is 1 if @var{name} is defined as a macro at
2887 the current point in the program, and 0 otherwise. Thus, @code{@w{#if
2888 defined MACRO}} is precisely equivalent to @code{@w{#ifdef MACRO}}.
2890 @code{defined} is useful when you wish to test more than one macro for
2891 existence at once. For example,
2894 #if defined (__vax__) || defined (__ns16000__)
2898 would succeed if either of the names @code{__vax__} or
2899 @code{__ns16000__} is defined as a macro.
2901 Conditionals written like this:
2904 #if defined BUFSIZE && BUFSIZE >= 1024
2908 can generally be simplified to just @code{@w{#if BUFSIZE >= 1024}},
2909 since if @code{BUFSIZE} is not defined, it will be interpreted as having
2912 If the @code{defined} operator appears as a result of a macro expansion,
2913 the C standard says the behavior is undefined. GNU cpp treats it as a
2914 genuine @code{defined} operator and evaluates it normally. It will warn
2915 wherever your code uses this feature if you use the command-line option
2916 @option{-pedantic}, since other compilers may handle it differently.
2922 The @samp{#else} directive can be added to a conditional to provide
2923 alternative text to be used if the condition fails. This is what it
2928 #if @var{expression}
2930 #else /* Not @var{expression} */
2932 #endif /* Not @var{expression} */
2937 If @var{expression} is nonzero, the @var{text-if-true} is included and
2938 the @var{text-if-false} is skipped. If @var{expression} is zero, the
2941 You can use @samp{#else} with @samp{#ifdef} and @samp{#ifndef}, too.
2947 One common case of nested conditionals is used to check for more than two
2948 possible alternatives. For example, you might have
2962 Another conditional directive, @samp{#elif}, allows this to be
2963 abbreviated as follows:
2970 #else /* X != 2 and X != 1*/
2972 #endif /* X != 2 and X != 1*/
2975 @samp{#elif} stands for ``else if''. Like @samp{#else}, it goes in the
2976 middle of a conditional group and subdivides it; it does not require a
2977 matching @samp{#endif} of its own. Like @samp{#if}, the @samp{#elif}
2978 directive includes an expression to be tested. The text following the
2979 @samp{#elif} is processed only if the original @samp{#if}-condition
2980 failed and the @samp{#elif} condition succeeds.
2982 More than one @samp{#elif} can go in the same conditional group. Then
2983 the text after each @samp{#elif} is processed only if the @samp{#elif}
2984 condition succeeds after the original @samp{#if} and all previous
2985 @samp{#elif} directives within it have failed.
2987 @samp{#else} is allowed after any number of @samp{#elif} directives, but
2988 @samp{#elif} may not follow @samp{#else}.
2991 @section Deleted Code
2992 @cindex commenting out code
2994 If you replace or delete a part of the program but want to keep the old
2995 code around for future reference, you often cannot simply comment it
2996 out. Block comments do not nest, so the first comment inside the old
2997 code will end the commenting-out. The probable result is a flood of
3000 One way to avoid this problem is to use an always-false conditional
3001 instead. For instance, put @code{#if 0} before the deleted code and
3002 @code{#endif} after it. This works even if the code being turned
3003 off contains conditionals, but they must be entire conditionals
3004 (balanced @samp{#if} and @samp{#endif}).
3006 Some people use @code{#ifdef notdef} instead. This is risky, because
3007 @code{notdef} might be accidentally defined as a macro, and then the
3008 conditional would succeed. @code{#if 0} can be counted on to fail.
3010 Do not use @code{#if 0} for comments which are not C code. Use a real
3011 comment, instead. The interior of @code{#if 0} must consist of complete
3012 tokens; in particular, single-quote characters must balance. Comments
3013 often contain unbalanced single-quote characters (known in English as
3014 apostrophes). These confuse @code{#if 0}. They don't confuse
3018 @chapter Diagnostics
3020 @cindex reporting errors
3021 @cindex reporting warnings
3024 The directive @samp{#error} causes the preprocessor to report a fatal
3025 error. The tokens forming the rest of the line following @samp{#error}
3026 are used as the error message.
3028 You would use @samp{#error} inside of a conditional that detects a
3029 combination of parameters which you know the program does not properly
3030 support. For example, if you know that the program will not run
3031 properly on a VAX, you might write
3036 #error "Won't work on VAXen. See comments at get_last_object."
3041 If you have several configuration parameters that must be set up by
3042 the installation in a consistent way, you can use conditionals to detect
3043 an inconsistency and report it with @samp{#error}. For example,
3046 #if !defined(UNALIGNED_INT_ASM_OP) && defined(DWARF2_DEBUGGING_INFO)
3047 #error "DWARF2_DEBUGGING_INFO requires UNALIGNED_INT_ASM_OP."
3052 The directive @samp{#warning} is like @samp{#error}, but causes the
3053 preprocessor to issue a warning and continue preprocessing. The tokens
3054 following @samp{#warning} are used as the warning message.
3056 You might use @samp{#warning} in obsolete header files, with a message
3057 directing the user to the header file which should be used instead.
3059 Neither @samp{#error} nor @samp{#warning} macro-expands its argument.
3060 Internal whitespace sequences are each replaced with a single space.
3061 The line must consist of complete tokens. It is wisest to make the
3062 argument of these directives be a single string constant; this avoids
3063 problems with apostrophes and the like.
3066 @chapter Line Control
3067 @cindex line control
3069 The C preprocessor informs the C compiler of the location in your source
3070 code where each token came from. Presently, this is just the file name
3071 and line number. All the tokens resulting from macro expansion are
3072 reported as having appeared on the line of the source file where the
3073 outermost macro was used. We intend to be more accurate in the future.
3075 If you write a program which generates source code, such as the
3076 @command{bison} parser generator, you may want to adjust the preprocessor's
3077 notion of the current file name and line number by hand. Parts of the
3078 output from @command{bison} are generated from scratch, other parts come
3079 from a standard parser file. The rest are copied verbatim from
3080 @command{bison}'s input. You would like compiler error messages and
3081 symbolic debuggers to be able to refer to @code{bison}'s input file.
3084 @command{bison} or any such program can arrange this by writing
3085 @samp{#line} directives into the output file. @samp{#line} is a
3086 directive that specifies the original line number and source file name
3087 for subsequent input in the current preprocessor input file.
3088 @samp{#line} has three variants:
3091 @item #line @var{linenum}
3092 @var{linenum} is a non-negative decimal integer constant. It specifies
3093 the line number which should be reported for the following line of
3094 input. Subsequent lines are counted from @var{linenum}.
3096 @item #line @var{linenum} @var{filename}
3097 @var{linenum} is the same as for the first form, and has the same
3098 effect. In addition, @var{filename} is a string constant. The
3099 following line and all subsequent lines are reported to come from the
3100 file it specifies, until something else happens to change that.
3101 @var{filename} is interpreted according to the normal rules for a string
3102 constant: backslash escapes are interpreted. This is different from
3105 Previous versions of CPP did not interpret escapes in @samp{#line};
3106 we have changed it because the standard requires they be interpreted,
3107 and most other compilers do.
3109 @item #line @var{anything else}
3110 @var{anything else} is checked for macro calls, which are expanded.
3111 The result should match one of the above two forms.
3114 @samp{#line} directives alter the results of the @code{__FILE__} and
3115 @code{__LINE__} predefined macros from that point on. @xref{Standard
3116 Predefined Macros}. They do not have any effect on @samp{#include}'s
3117 idea of the directory containing the current file. This is a change
3118 from GCC 2.95. Previously, a file reading
3121 #line 1 "../src/gram.y"
3125 would search for @file{gram.h} in @file{../src}, then the @option{-I}
3126 chain; the directory containing the physical source file would not be
3127 searched. In GCC 3.0 and later, the @samp{#include} is not affected by
3128 the presence of a @samp{#line} referring to a different directory.
3130 We made this change because the old behavior caused problems when
3131 generated source files were transported between machines. For instance,
3132 it is common practice to ship generated parsers with a source release,
3133 so that people building the distribution do not need to have yacc or
3134 Bison installed. These files frequently have @samp{#line} directives
3135 referring to the directory tree of the system where the distribution was
3136 created. If GCC tries to search for headers in those directories, the
3137 build is likely to fail.
3139 The new behavior can cause failures too, if the generated file is not
3140 in the same directory as its source and it attempts to include a header
3141 which would be visible searching from the directory containing the
3142 source file. However, this problem is easily solved with an additional
3143 @option{-I} switch on the command line. The failures caused by the old
3144 semantics could sometimes be corrected only by editing the generated
3145 files, which is difficult and error-prone.
3150 The @samp{#pragma} directive is the method specified by the C standard
3151 for providing additional information to the compiler, beyond what is
3152 conveyed in the language itself. Three forms of this directive
3153 (commonly known as @dfn{pragmas}) are specified by the 1999 C standard.
3154 A C compiler is free to attach any meaning it likes to other pragmas.
3156 GCC has historically preferred to use extensions to the syntax of the
3157 language, such as @code{__attribute__}, for this purpose. However, GCC
3158 does define a few pragmas of its own. These mostly have effects on the
3159 entire translation unit or source file.
3161 In GCC version 3, all GNU-defined, supported pragmas have been given a
3162 @code{GCC} prefix. This is in line with the @code{STDC} prefix on all
3163 pragmas defined by C99. For backward compatibility, pragmas which were
3164 recognized by previous versions are still recognized without the
3165 @code{GCC} prefix, but that usage is deprecated. Some older pragmas are
3166 deprecated in their entirety. They are not recognized with the
3167 @code{GCC} prefix. @xref{Obsolete Features}.
3169 @cindex @code{_Pragma}
3170 C99 introduces the @code{@w{_Pragma}} operator. This feature addresses a
3171 major problem with @samp{#pragma}: being a directive, it cannot be
3172 produced as the result of macro expansion. @code{@w{_Pragma}} is an
3173 operator, much like @code{sizeof} or @code{defined}, and can be embedded
3176 Its syntax is @code{@w{_Pragma (@var{string-literal})}}, where
3177 @var{string-literal} can be either a normal or wide-character string
3178 literal. It is destringized, by replacing all @samp{\\} with a single
3179 @samp{\} and all @samp{\"} with a @samp{"}. The result is then
3180 processed as if it had appeared as the right hand side of a
3181 @samp{#pragma} directive. For example,
3184 _Pragma ("GCC dependency \"parse.y\"")
3188 has the same effect as @code{#pragma GCC dependency "parse.y"}. The
3189 same effect could be achieved using macros, for example
3192 #define DO_PRAGMA(x) _Pragma (#x)
3193 DO_PRAGMA (GCC dependency "parse.y")
3196 The standard is unclear on where a @code{_Pragma} operator can appear.
3197 The preprocessor does not accept it within a preprocessing conditional
3198 directive like @samp{#if}. To be safe, you are probably best keeping it
3199 out of directives other than @samp{#define}, and putting it on a line of
3202 This manual documents the pragmas which are meaningful to the
3203 preprocessor itself. Other pragmas are meaningful to the C or C++
3204 compilers. They are documented in the GCC manual.
3207 @item #pragma GCC dependency
3208 @code{#pragma GCC dependency} allows you to check the relative dates of
3209 the current file and another file. If the other file is more recent than
3210 the current file, a warning is issued. This is useful if the current
3211 file is derived from the other file, and should be regenerated. The
3212 other file is searched for using the normal include search path.
3213 Optional trailing text can be used to give more information in the
3217 #pragma GCC dependency "parse.y"
3218 #pragma GCC dependency "/usr/include/time.h" rerun fixincludes
3221 @item #pragma GCC poison
3222 Sometimes, there is an identifier that you want to remove completely
3223 from your program, and make sure that it never creeps back in. To
3224 enforce this, you can @dfn{poison} the identifier with this pragma.
3225 @code{#pragma GCC poison} is followed by a list of identifiers to
3226 poison. If any of those identifiers appears anywhere in the source
3227 after the directive, it is a hard error. For example,
3230 #pragma GCC poison printf sprintf fprintf
3231 sprintf(some_string, "hello");
3235 will produce an error.
3237 If a poisoned identifier appears as part of the expansion of a macro
3238 which was defined before the identifier was poisoned, it will @emph{not}
3239 cause an error. This lets you poison an identifier without worrying
3240 about system headers defining macros that use it.
3245 #define strrchr rindex
3246 #pragma GCC poison rindex
3247 strrchr(some_string, 'h');
3251 will not produce an error.
3253 @item #pragma GCC system_header
3254 This pragma takes no arguments. It causes the rest of the code in the
3255 current file to be treated as if it came from a system header.
3256 @xref{System Headers}.
3260 @node Other Directives
3261 @chapter Other Directives
3264 The @samp{#ident} directive takes one argument, a string constant. On
3265 some systems, that string constant is copied into a special segment of
3266 the object file. On other systems, the directive is ignored.
3268 This directive is not part of the C standard, but it is not an official
3269 GNU extension either. We believe it came from System V@.
3272 The @samp{#sccs} directive is recognized, because it appears in the
3273 header files of some systems. It is a very old, obscure, extension
3274 which we did not invent, and we have been unable to find any
3275 documentation of what it should do, so GCC simply ignores it.
3277 @cindex null directive
3278 The @dfn{null directive} consists of a @samp{#} followed by a newline,
3279 with only whitespace (including comments) in between. A null directive
3280 is understood as a preprocessing directive but has no effect on the
3281 preprocessor output. The primary significance of the existence of the
3282 null directive is that an input line consisting of just a @samp{#} will
3283 produce no output, rather than a line of output containing just a
3284 @samp{#}. Supposedly some old C programs contain such lines.
3286 @node Preprocessor Output
3287 @chapter Preprocessor Output
3289 When the C preprocessor is used with the C, C++, or Objective-C
3290 compilers, it is integrated into the compiler and communicates a stream
3291 of binary tokens directly to the compiler's parser. However, it can
3292 also be used in the more conventional standalone mode, where it produces
3294 @c FIXME: Document the library interface.
3296 @cindex output format
3297 The output from the C preprocessor looks much like the input, except
3298 that all preprocessing directive lines have been replaced with blank
3299 lines and all comments with spaces. Long runs of blank lines are
3302 The ISO standard specifies that it is implementation defined whether a
3303 preprocessor preserves whitespace between tokens, or replaces it with
3304 e.g.@: a single space. In GNU CPP, whitespace between tokens is collapsed
3305 to become a single space, with the exception that the first token on a
3306 non-directive line is preceded with sufficient spaces that it appears in
3307 the same column in the preprocessed output that it appeared in the
3308 original source file. This is so the output is easy to read.
3309 @xref{Differences from previous versions}. CPP does not insert any
3310 whitespace where there was none in the original source, except where
3311 necessary to prevent an accidental token paste.
3314 Source file name and line number information is conveyed by lines
3318 # @var{linenum} @var{filename} @var{flags}
3322 These are called @dfn{linemarkers}. They are inserted as needed into
3323 the output (but never within a string or character constant). They mean
3324 that the following line originated in file @var{filename} at line
3325 @var{linenum}. @var{filename} will never contain any non-printing
3326 characters; they are replaced with octal escape sequences.
3328 After the file name comes zero or more flags, which are @samp{1},
3329 @samp{2}, @samp{3}, or @samp{4}. If there are multiple flags, spaces
3330 separate them. Here is what the flags mean:
3334 This indicates the start of a new file.
3336 This indicates returning to a file (after having included another file).
3338 This indicates that the following text comes from a system header file,
3339 so certain warnings should be suppressed.
3341 This indicates that the following text should be treated as being
3342 wrapped in an implicit @code{extern "C"} block.
3343 @c maybe cross reference NO_IMPLICIT_EXTERN_C
3346 As an extension, the preprocessor accepts linemarkers in non-assembler
3347 input files. They are treated like the corresponding @samp{#line}
3348 directive, (@pxref{Line Control}), except that trailing flags are
3349 permitted, and are interpreted with the meanings described above. If
3350 multiple flags are given, they must be in ascending order.
3352 Some directives may be duplicated in the output of the preprocessor.
3353 These are @samp{#ident} (always), @samp{#pragma} (only if the
3354 preprocessor does not handle the pragma itself), and @samp{#define} and
3355 @samp{#undef} (with certain debugging options). If this happens, the
3356 @samp{#} of the directive will always be in the first column, and there
3357 will be no space between the @samp{#} and the directive name. If macro
3358 expansion happens to generate tokens which might be mistaken for a
3359 duplicated directive, a space will be inserted between the @samp{#} and
3362 @node Traditional Mode
3363 @chapter Traditional Mode
3365 Traditional (pre-standard) C preprocessing is rather different from
3366 the preprocessing specified by the standard. When GCC is given the
3367 @option{-traditional-cpp} option, it attempts to emulate a traditional
3370 GCC versions 3.2 and later only support traditional mode semantics in
3371 the preprocessor, and not in the compiler front ends. This chapter
3372 outlines the traditional preprocessor semantics we implemented.
3374 The implementation does not correspond precisely to the behavior of
3375 earlier versions of GCC, nor to any true traditional preprocessor.
3376 After all, inconsistencies among traditional implementations were a
3377 major motivation for C standardization. However, we intend that it
3378 should be compatible with true traditional preprocessors in all ways
3379 that actually matter.
3382 * Traditional lexical analysis::
3383 * Traditional macros::
3384 * Traditional miscellany::
3385 * Traditional warnings::
3388 @node Traditional lexical analysis
3389 @section Traditional lexical analysis
3391 The traditional preprocessor does not decompose its input into tokens
3392 the same way a standards-conforming preprocessor does. The input is
3393 simply treated as a stream of text with minimal internal form.
3395 This implementation does not treat trigraphs (@pxref{trigraphs})
3396 specially since they were an invention of the standards committee. It
3397 handles arbitrarily-positioned escaped newlines properly and splices
3398 the lines as you would expect; many traditional preprocessors did not
3401 The form of horizontal whitespace in the input file is preserved in
3402 the output. In particular, hard tabs remain hard tabs. This can be
3403 useful if, for example, you are preprocessing a Makefile.
3405 Traditional CPP only recognizes C-style block comments, and treats the
3406 @samp{/*} sequence as introducing a comment only if it lies outside
3407 quoted text. Quoted text is introduced by the usual single and double
3408 quotes, and also by an initial @samp{<} in a @code{#include}
3411 Traditionally, comments are completely removed and are not replaced
3412 with a space. Since a traditional compiler does its own tokenization
3413 of the output of the preprocessor, this means that comments can
3414 effectively be used as token paste operators. However, comments
3415 behave like separators for text handled by the preprocessor itself,
3416 since it doesn't re-lex its input. For example, in
3423 @samp{foo} and @samp{bar} are distinct identifiers and expanded
3424 separately if they happen to be macros. In other words, this
3425 directive is equivalent to
3438 Generally speaking, in traditional mode an opening quote need not have
3439 a matching closing quote. In particular, a macro may be defined with
3440 replacement text that contains an unmatched quote. Of course, if you
3441 attempt to compile preprocessed output containing an unmatched quote
3442 you will get a syntax error.
3444 However, all preprocessing directives other than @code{#define}
3445 require matching quotes. For example:
3448 #define m This macro's fine and has an unmatched quote
3449 "/* This is not a comment. */
3450 /* This is a comment. The following #include directive
3455 Just as for the ISO preprocessor, what would be a closing quote can be
3456 escaped with a backslash to prevent the quoted text from closing.
3458 @node Traditional macros
3459 @section Traditional macros
3461 The major difference between traditional and ISO macros is that the
3462 former expand to text rather than to a token sequence. CPP removes
3463 all leading and trailing horizontal whitespace from a macro's
3464 replacement text before storing it, but preserves the form of internal
3467 One consequence is that it is legitimate for the replacement text to
3468 contain an unmatched quote (@pxref{Traditional lexical analysis}). An
3469 unclosed string or character constant continues into the text
3470 following the macro call. Similarly, the text at the end of a macro's
3471 expansion can run together with the text after the macro invocation to
3472 produce a single token.
3474 Normally comments are removed from the replacement text after the
3475 macro is expanded, but if the @option{-CC} option is passed on the
3476 command line comments are preserved. (In fact, the current
3477 implementation removes comments even before saving the macro
3478 replacement text, but it careful to do it in such a way that the
3479 observed effect is identical even in the function-like macro case.)
3481 The ISO stringification operator @samp{#} and token paste operator
3482 @samp{##} have no special meaning. As explained later, an effect
3483 similar to these operators can be obtained in a different way. Macro
3484 names that are embedded in quotes, either from the main file or after
3485 macro replacement, do not expand.
3487 CPP replaces an unquoted object-like macro name with its replacement
3488 text, and then rescans it for further macros to replace. Unlike
3489 standard macro expansion, traditional macro expansion has no provision
3490 to prevent recursion. If an object-like macro appears unquoted in its
3491 replacement text, it will be replaced again during the rescan pass,
3492 and so on @emph{ad infinitum}. GCC detects when it is expanding
3493 recursive macros, emits an error message, and continues after the
3494 offending macro invocation.
3498 #define INC(x) PLUS+x
3503 Function-like macros are similar in form but quite different in
3504 behavior to their ISO counterparts. Their arguments are contained
3505 within parentheses, are comma-separated, and can cross physical lines.
3506 Commas within nested parentheses are not treated as argument
3507 separators. Similarly, a quote in an argument cannot be left
3508 unclosed; a following comma or parenthesis that comes before the
3509 closing quote is treated like any other character. There is no
3510 facility for handling variadic macros.
3512 This implementation removes all comments from macro arguments, unless
3513 the @option{-C} option is given. The form of all other horizontal
3514 whitespace in arguments is preserved, including leading and trailing
3515 whitespace. In particular
3522 is treated as an invocation of the macro @samp{f} with a single
3523 argument consisting of a single space. If you want to invoke a
3524 function-like macro that takes no arguments, you must not leave any
3525 whitespace between the parentheses.
3527 If a macro argument crosses a new line, the new line is replaced with
3528 a space when forming the argument. If the previous line contained an
3529 unterminated quote, the following line inherits the quoted state.
3531 Traditional preprocessors replace parameters in the replacement text
3532 with their arguments regardless of whether the parameters are within
3533 quotes or not. This provides a way to stringize arguments. For
3538 str(/* A comment */some text )
3539 @expansion{} "some text "
3543 Note that the comment is removed, but that the trailing space is
3544 preserved. Here is an example of using a comment to effect token
3548 #define suffix(x) foo_/**/x
3550 @expansion{} foo_bar
3553 @node Traditional miscellany
3554 @section Traditional miscellany
3556 Here are some things to be aware of when using the traditional
3561 Preprocessing directives are recognized only when their leading
3562 @samp{#} appears in the first column. There can be no whitespace
3563 between the beginning of the line and the @samp{#}, but whitespace can
3564 follow the @samp{#}.
3567 A true traditional C preprocessor does not recognize @samp{#error} or
3568 @samp{#pragma}, and may not recognize @samp{#elif}. CPP supports all
3569 the directives in traditional mode that it supports in ISO mode,
3570 including extensions, with the exception that the effects of
3571 @samp{#pragma GCC poison} are undefined.
3574 __STDC__ is not defined.
3577 If you use digraphs the behaviour is undefined.
3580 If a line that looks like a directive appears within macro arguments,
3581 the behaviour is undefined.
3585 @node Traditional warnings
3586 @section Traditional warnings
3587 You can request warnings about features that did not exist, or worked
3588 differently, in traditional C with the @option{-Wtraditional} option.
3589 GCC does not warn about features of ISO C which you must use when you
3590 are using a conforming compiler, such as the @samp{#} and @samp{##}
3593 Presently @option{-Wtraditional} warns about:
3597 Macro parameters that appear within string literals in the macro body.
3598 In traditional C macro replacement takes place within string literals,
3599 but does not in ISO C@.
3602 In traditional C, some preprocessor directives did not exist.
3603 Traditional preprocessors would only consider a line to be a directive
3604 if the @samp{#} appeared in column 1 on the line. Therefore
3605 @option{-Wtraditional} warns about directives that traditional C
3606 understands but would ignore because the @samp{#} does not appear as the
3607 first character on the line. It also suggests you hide directives like
3608 @samp{#pragma} not understood by traditional C by indenting them. Some
3609 traditional implementations would not recognize @samp{#elif}, so it
3610 suggests avoiding it altogether.
3613 A function-like macro that appears without an argument list. In some
3614 traditional preprocessors this was an error. In ISO C it merely means
3615 that the macro is not expanded.
3618 The unary plus operator. This did not exist in traditional C@.
3621 The @samp{U} and @samp{LL} integer constant suffixes, which were not
3622 available in traditional C@. (Traditional C does support the @samp{L}
3623 suffix for simple long integer constants.) You are not warned about
3624 uses of these suffixes in macros defined in system headers. For
3625 instance, @code{UINT_MAX} may well be defined as @code{4294967295U}, but
3626 you will not be warned if you use @code{UINT_MAX}.
3628 You can usually avoid the warning, and the related warning about
3629 constants which are so large that they are unsigned, by writing the
3630 integer constant in question in hexadecimal, with no U suffix. Take
3631 care, though, because this gives the wrong result in exotic cases.
3634 @node Implementation Details
3635 @chapter Implementation Details
3637 Here we document details of how the preprocessor's implementation
3638 affects its user-visible behavior. You should try to avoid undue
3639 reliance on behavior described here, as it is possible that it will
3640 change subtly in future implementations.
3642 Also documented here are obsolete features and changes from previous
3646 * Implementation-defined behavior::
3647 * Implementation limits::
3648 * Obsolete Features::
3649 * Differences from previous versions::
3652 @node Implementation-defined behavior
3653 @section Implementation-defined behavior
3654 @cindex implementation-defined behavior
3656 This is how CPP behaves in all the cases which the C standard
3657 describes as @dfn{implementation-defined}. This term means that the
3658 implementation is free to do what it likes, but must document its choice
3660 @c FIXME: Check the C++ standard for more implementation-defined stuff.
3664 @item The mapping of physical source file multi-byte characters to the
3665 execution character set.
3667 Currently, GNU cpp only supports character sets that are strict supersets
3668 of ASCII, and performs no translation of characters.
3670 @item Non-empty sequences of whitespace characters.
3672 In textual output, each whitespace sequence is collapsed to a single
3673 space. For aesthetic reasons, the first token on each non-directive
3674 line of output is preceded with sufficient spaces that it appears in the
3675 same column as it did in the original source file.
3677 @item The numeric value of character constants in preprocessor expressions.
3679 The preprocessor and compiler interpret character constants in the
3680 same way; i.e.@: escape sequences such as @samp{\a} are given the
3681 values they would have on the target machine.
3683 The compiler values a multi-character character constant a character
3684 at a time, shifting the previous value left by the number of bits per
3685 target character, and then or-ing in the bit-pattern of the new
3686 character truncated to the width of a target character. The final
3687 bit-pattern is given type @code{int}, and is therefore signed,
3688 regardless of whether single characters are signed or not (a slight
3689 change from versions 3.1 and earlier of GCC). If there are more
3690 characters in the constant than would fit in the target @code{int} the
3691 compiler issues a warning, and the excess leading characters are
3694 For example, 'ab' for a target with an 8-bit @code{char} would be
3695 interpreted as @w{(int) ((unsigned char) 'a' * 256 + (unsigned char)
3696 'b')}, and '\234a' as @w{(int) ((unsigned char) '\234' * 256 + (unsigned
3699 @item Source file inclusion.
3701 For a discussion on how the preprocessor locates header files,
3702 @ref{Include Operation}.
3704 @item Interpretation of the filename resulting from a macro-expanded
3705 @samp{#include} directive.
3707 @xref{Computed Includes}.
3709 @item Treatment of a @samp{#pragma} directive that after macro-expansion
3710 results in a standard pragma.
3712 No macro expansion occurs on any @samp{#pragma} directive line, so the
3713 question does not arise.
3715 Note that GCC does not yet implement any of the standard
3720 @node Implementation limits
3721 @section Implementation limits
3722 @cindex implementation limits
3724 CPP has a small number of internal limits. This section lists the
3725 limits which the C standard requires to be no lower than some minimum,
3726 and all the others we are aware of. We intend there to be as few limits
3727 as possible. If you encounter an undocumented or inconvenient limit,
3728 please report that to us as a bug. (See the section on reporting bugs in
3731 Where we say something is limited @dfn{only by available memory}, that
3732 means that internal data structures impose no intrinsic limit, and space
3733 is allocated with @code{malloc} or equivalent. The actual limit will
3734 therefore depend on many things, such as the size of other things
3735 allocated by the compiler at the same time, the amount of memory
3736 consumed by other processes on the same computer, etc.
3740 @item Nesting levels of @samp{#include} files.
3742 We impose an arbitrary limit of 200 levels, to avoid runaway recursion.
3743 The standard requires at least 15 levels.
3745 @item Nesting levels of conditional inclusion.
3747 The C standard mandates this be at least 63. CPP is limited only by
3750 @item Levels of parenthesised expressions within a full expression.
3752 The C standard requires this to be at least 63. In preprocessor
3753 conditional expressions, it is limited only by available memory.
3755 @item Significant initial characters in an identifier or macro name.
3757 The preprocessor treats all characters as significant. The C standard
3758 requires only that the first 63 be significant.
3760 @item Number of macros simultaneously defined in a single translation unit.
3762 The standard requires at least 4095 be possible. CPP is limited only
3763 by available memory.
3765 @item Number of parameters in a macro definition and arguments in a macro call.
3767 We allow @code{USHRT_MAX}, which is no smaller than 65,535. The minimum
3768 required by the standard is 127.
3770 @item Number of characters on a logical source line.
3772 The C standard requires a minimum of 4096 be permitted. CPP places
3773 no limits on this, but you may get incorrect column numbers reported in
3774 diagnostics for lines longer than 65,535 characters.
3776 @item Maximum size of a source file.
3778 The standard does not specify any lower limit on the maximum size of a
3779 source file. GNU cpp maps files into memory, so it is limited by the
3780 available address space. This is generally at least two gigabytes.
3781 Depending on the operating system, the size of physical memory may or
3782 may not be a limitation.
3786 @node Obsolete Features
3787 @section Obsolete Features
3789 CPP has a number of features which are present mainly for
3790 compatibility with older programs. We discourage their use in new code.
3791 In some cases, we plan to remove the feature in a future version of GCC@.
3795 * Obsolete once-only headers::
3799 @subsection Assertions
3802 @dfn{Assertions} are a deprecated alternative to macros in writing
3803 conditionals to test what sort of computer or system the compiled
3804 program will run on. Assertions are usually predefined, but you can
3805 define them with preprocessing directives or command-line options.
3807 Assertions were intended to provide a more systematic way to describe
3808 the compiler's target system. However, in practice they are just as
3809 unpredictable as the system-specific predefined macros. In addition, they
3810 are not part of any standard, and only a few compilers support them.
3811 Therefore, the use of assertions is @strong{less} portable than the use
3812 of system-specific predefined macros. We recommend you do not use them at
3816 An assertion looks like this:
3819 #@var{predicate} (@var{answer})
3823 @var{predicate} must be a single identifier. @var{answer} can be any
3824 sequence of tokens; all characters are significant except for leading
3825 and trailing whitespace, and differences in internal whitespace
3826 sequences are ignored. (This is similar to the rules governing macro
3827 redefinition.) Thus, @code{(x + y)} is different from @code{(x+y)} but
3828 equivalent to @code{@w{( x + y )}}. Parentheses do not nest inside an
3831 @cindex testing predicates
3832 To test an assertion, you write it in an @samp{#if}. For example, this
3833 conditional succeeds if either @code{vax} or @code{ns16000} has been
3834 asserted as an answer for @code{machine}.
3837 #if #machine (vax) || #machine (ns16000)
3841 You can test whether @emph{any} answer is asserted for a predicate by
3842 omitting the answer in the conditional:
3849 Assertions are made with the @samp{#assert} directive. Its sole
3850 argument is the assertion to make, without the leading @samp{#} that
3851 identifies assertions in conditionals.
3854 #assert @var{predicate} (@var{answer})
3858 You may make several assertions with the same predicate and different
3859 answers. Subsequent assertions do not override previous ones for the
3860 same predicate. All the answers for any given predicate are
3861 simultaneously true.
3863 @cindex assertions, cancelling
3865 Assertions can be cancelled with the @samp{#unassert} directive. It
3866 has the same syntax as @samp{#assert}. In that form it cancels only the
3867 answer which was specified on the @samp{#unassert} line; other answers
3868 for that predicate remain true. You can cancel an entire predicate by
3869 leaving out the answer:
3872 #unassert @var{predicate}
3876 In either form, if no such assertion has been made, @samp{#unassert} has
3879 You can also make or cancel assertions using command line options.
3882 @node Obsolete once-only headers
3883 @subsection Obsolete once-only headers
3885 CPP supports two more ways of indicating that a header file should be
3886 read only once. Neither one is as portable as a wrapper @samp{#ifndef},
3887 and we recommend you do not use them in new programs.
3890 In the Objective-C language, there is a variant of @samp{#include}
3891 called @samp{#import} which includes a file, but does so at most once.
3892 If you use @samp{#import} instead of @samp{#include}, then you don't
3893 need the conditionals inside the header file to prevent multiple
3894 inclusion of the contents. GCC permits the use of @samp{#import} in C
3895 and C++ as well as Objective-C@. However, it is not in standard C or C++
3896 and should therefore not be used by portable programs.
3898 @samp{#import} is not a well designed feature. It requires the users of
3899 a header file to know that it should only be included once. It is much
3900 better for the header file's implementor to write the file so that users
3901 don't need to know this. Using a wrapper @samp{#ifndef} accomplishes
3904 In the present implementation, a single use of @samp{#import} will
3905 prevent the file from ever being read again, by either @samp{#import} or
3906 @samp{#include}. You should not rely on this; do not use both
3907 @samp{#import} and @samp{#include} to refer to the same header file.
3909 Another way to prevent a header file from being included more than once
3910 is with the @samp{#pragma once} directive. If @samp{#pragma once} is
3911 seen when scanning a header file, that file will never be read again, no
3914 @samp{#pragma once} does not have the problems that @samp{#import} does,
3915 but it is not recognized by all preprocessors, so you cannot rely on it
3916 in a portable program.
3918 @node Differences from previous versions
3919 @section Differences from previous versions
3920 @cindex differences from previous versions
3922 This section details behavior which has changed from previous versions
3923 of CPP@. We do not plan to change it again in the near future, but
3924 we do not promise not to, either.
3926 The ``previous versions'' discussed here are 2.95 and before. The
3927 behavior of GCC 3.0 is mostly the same as the behavior of the widely
3928 used 2.96 and 2.97 development snapshots. Where there are differences,
3929 they generally represent bugs in the snapshots.
3933 @item Order of evaluation of @samp{#} and @samp{##} operators
3935 The standard does not specify the order of evaluation of a chain of
3936 @samp{##} operators, nor whether @samp{#} is evaluated before, after, or
3937 at the same time as @samp{##}. You should therefore not write any code
3938 which depends on any specific ordering. It is possible to guarantee an
3939 ordering, if you need one, by suitable use of nested macros.
3941 An example of where this might matter is pasting the arguments @samp{1},
3942 @samp{e} and @samp{-2}. This would be fine for left-to-right pasting,
3943 but right-to-left pasting would produce an invalid token @samp{e-2}.
3945 GCC 3.0 evaluates @samp{#} and @samp{##} at the same time and strictly
3946 left to right. Older versions evaluated all @samp{#} operators first,
3947 then all @samp{##} operators, in an unreliable order.
3949 @item The form of whitespace betwen tokens in preprocessor output
3951 @xref{Preprocessor Output}, for the current textual format. This is
3952 also the format used by stringification. Normally, the preprocessor
3953 communicates tokens directly to the compiler's parser, and whitespace
3954 does not come up at all.
3956 Older versions of GCC preserved all whitespace provided by the user and
3957 inserted lots more whitespace of their own, because they could not
3958 accurately predict when extra spaces were needed to prevent accidental
3961 @item Optional argument when invoking rest argument macros
3963 As an extension, GCC permits you to omit the variable arguments entirely
3964 when you use a variable argument macro. This is forbidden by the 1999 C
3965 standard, and will provoke a pedantic warning with GCC 3.0. Previous
3966 versions accepted it silently.
3968 @item @samp{##} swallowing preceding text in rest argument macros
3970 Formerly, in a macro expansion, if @samp{##} appeared before a variable
3971 arguments parameter, and the set of tokens specified for that argument
3972 in the macro invocation was empty, previous versions of CPP would
3973 back up and remove the preceding sequence of non-whitespace characters
3974 (@strong{not} the preceding token). This extension is in direct
3975 conflict with the 1999 C standard and has been drastically pared back.
3977 In the current version of the preprocessor, if @samp{##} appears between
3978 a comma and a variable arguments parameter, and the variable argument is
3979 omitted entirely, the comma will be removed from the expansion. If the
3980 variable argument is empty, or the token before @samp{##} is not a
3981 comma, then @samp{##} behaves as a normal token paste.
3983 @item @samp{#line} and @samp{#include}
3985 The @samp{#line} directive used to change GCC's notion of the
3986 ``directory containing the current file,'' used by @samp{#include} with
3987 a double-quoted header file name. In 3.0 and later, it does not.
3988 @xref{Line Control}, for further explanation.
3990 @item Syntax of @samp{#line}
3992 In GCC 2.95 and previous, the string constant argument to @samp{#line}
3993 was treated the same way as the argument to @samp{#include}: backslash
3994 escapes were not honored, and the string ended at the second @samp{"}.
3995 This is not compliant with the C standard. In GCC 3.0, an attempt was
3996 made to correct the behavior, so that the string was treated as a real
3997 string constant, but it turned out to be buggy. In 3.1, the bugs have
3998 been fixed. (We are not fixing the bugs in 3.0 because they affect
3999 relatively few people and the fix is quite invasive.)
4006 @cindex command line
4008 Most often when you use the C preprocessor you will not have to invoke it
4009 explicitly: the C compiler will do so automatically. However, the
4010 preprocessor is sometimes useful on its own. All the options listed
4011 here are also acceptable to the C compiler and have the same meaning,
4012 except that the C compiler has different rules for specifying the output
4015 @strong{Note:} Whether you use the preprocessor by way of @command{gcc}
4016 or @command{cpp}, the @dfn{compiler driver} is run first. This
4017 program's purpose is to translate your command into invocations of the
4018 programs that do the actual work. Their command line interfaces are
4019 similar but not identical to the documented interface, and may change
4023 @c man begin SYNOPSIS
4024 cpp [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
4025 [@option{-I}@var{dir}@dots{}] [@option{-W}@var{warn}@dots{}]
4026 [@option{-M}|@option{-MM}] [@option{-MG}] [@option{-MF} @var{filename}]
4027 [@option{-MP}] [@option{-MQ} @var{target}@dots{}] [@option{-MT} @var{target}@dots{}]
4028 [@option{-x} @var{language}] [@option{-std=}@var{standard}]
4029 @var{infile} @var{outfile}
4031 Only the most useful options are listed here; see below for the remainder.
4033 @c man begin SEEALSO
4034 gpl(7), gfdl(7), fsf-funding(7),
4035 gcc(1), as(1), ld(1), and the Info entries for @file{cpp}, @file{gcc}, and
4040 @c man begin OPTIONS
4041 The C preprocessor expects two file names as arguments, @var{infile} and
4042 @var{outfile}. The preprocessor reads @var{infile} together with any
4043 other files it specifies with @samp{#include}. All the output generated
4044 by the combined input files is written in @var{outfile}.
4046 Either @var{infile} or @var{outfile} may be @option{-}, which as
4047 @var{infile} means to read from standard input and as @var{outfile}
4048 means to write to standard output. Also, if either file is omitted, it
4049 means the same as if @option{-} had been specified for that file.
4051 Unless otherwise noted, or the option ends in @samp{=}, all options
4052 which take an argument may have that argument appear either immediately
4053 after the option, or with a space between option and argument:
4054 @option{-Ifoo} and @option{-I foo} have the same effect.
4056 @cindex grouping options
4057 @cindex options, grouping
4058 Many options have multi-letter names; therefore multiple single-letter
4059 options may @emph{not} be grouped: @option{-dM} is very different from
4063 @include cppopts.texi
4066 @node Environment Variables
4067 @chapter Environment Variables
4068 @cindex environment variables
4069 @c man begin ENVIRONMENT
4071 This section describes the environment variables that affect how CPP
4072 operates. You can use them to specify directories or prefixes to use
4073 when searching for include files, or to control dependency output.
4075 Note that you can also specify places to search using options such as
4076 @option{-I}, and control dependency output with options like
4077 @option{-M} (@pxref{Invocation}). These take precedence over
4078 environment variables, which in turn take precedence over the
4079 configuration of GCC@.
4081 @include cppenv.texi
4088 @node Index of Directives
4089 @unnumbered Index of Directives
4093 @unnumbered Option Index
4095 CPP's command line options and environment variables are indexed here
4096 without any initial @samp{-} or @samp{--}.
4101 @unnumbered Concept Index